2012-11-29 13:28:09 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2012-12-19 22:19:30 +01:00
|
|
|
|
2015-02-22 08:58:50 -08:00
|
|
|
|
2016-06-30 18:40:10 -07:00
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
2017-02-02 19:15:33 +01:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-12-17 20:04:08 -08:00
|
|
|
|
2013-04-23 16:38:02 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
f2fs: don't wait writeback for datas during checkpoint
Normally, while committing checkpoint, we will wait on all pages to be
writebacked no matter the page is data or metadata, so in scenario where
there are lots of data IO being submitted with metadata, we may suffer
long latency for waiting writeback during checkpoint.
Indeed, we only care about persistence for pages with metadata, but not
pages with data, as file system consistent are only related to metadate,
so in order to avoid encountering long latency in above scenario, let's
recognize and reference metadata in submitted IOs, wait writeback only
for metadatas.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-11-16 10:41:20 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-07-20 15:29:37 +02:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2014-01-30 11:19:05 -08:00
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2016-09-18 23:30:07 +08:00
|
|
|
|
2017-02-25 11:08:28 +08:00
|
|
|
|
|
|
|
|
|
2017-06-03 09:38:06 +02:00
|
|
|
|
2017-02-25 11:08:28 +08:00
|
|
|
|
2016-09-18 23:30:07 +08:00
|
|
|
|
|
|
|
|
|
2015-05-25 18:03:38 +08:00
|
|
|
|
2017-06-03 09:38:06 +02:00
|
|
|
|
2015-05-15 16:26:10 -07:00
|
|
|
|
2015-05-25 18:03:38 +08:00
|
|
|
|
2015-05-15 16:26:10 -07:00
|
|
|
|
2015-05-25 18:03:38 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-01-30 11:19:05 -08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2017-06-03 09:38:06 +02:00
|
|
|
|
2016-06-30 18:49:15 -07:00
|
|
|
|
|
|
|
|
|
2014-01-30 11:19:05 -08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-01-30 11:19:05 -08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-07-20 15:29:37 +02:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2014-02-03 10:50:22 +09:00
|
|
|
|
2014-01-30 11:19:05 -08:00
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2014-01-30 11:19:05 -08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
f2fs: don't wait writeback for datas during checkpoint
Normally, while committing checkpoint, we will wait on all pages to be
writebacked no matter the page is data or metadata, so in scenario where
there are lots of data IO being submitted with metadata, we may suffer
long latency for waiting writeback during checkpoint.
Indeed, we only care about persistence for pages with metadata, but not
pages with data, as file system consistent are only related to metadate,
so in order to avoid encountering long latency in above scenario, let's
recognize and reference metadata in submitted IOs, wait writeback only
for metadatas.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-11-16 10:41:20 +08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-03 09:38:06 +02:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-05-15 16:26:10 -07:00
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2017-06-03 09:38:06 +02:00
|
|
|
|
2016-10-11 13:56:01 -07:00
|
|
|
|
2016-05-18 14:07:56 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
f2fs: don't wait writeback for datas during checkpoint
Normally, while committing checkpoint, we will wait on all pages to be
writebacked no matter the page is data or metadata, so in scenario where
there are lots of data IO being submitted with metadata, we may suffer
long latency for waiting writeback during checkpoint.
Indeed, we only care about persistence for pages with metadata, but not
pages with data, as file system consistent are only related to metadate,
so in order to avoid encountering long latency in above scenario, let's
recognize and reference metadata in submitted IOs, wait writeback only
for metadatas.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-11-16 10:41:20 +08:00
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2014-01-30 11:19:05 -08:00
|
|
|
|
f2fs: don't wait writeback for datas during checkpoint
Normally, while committing checkpoint, we will wait on all pages to be
writebacked no matter the page is data or metadata, so in scenario where
there are lots of data IO being submitted with metadata, we may suffer
long latency for waiting writeback during checkpoint.
Indeed, we only care about persistence for pages with metadata, but not
pages with data, as file system consistent are only related to metadate,
so in order to avoid encountering long latency in above scenario, let's
recognize and reference metadata in submitted IOs, wait writeback only
for metadatas.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-11-16 10:41:20 +08:00
|
|
|
|
2016-05-17 16:23:36 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-10-06 19:02:05 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-23 19:10:32 +02:00
|
|
|
|
2016-10-06 19:02:05 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-23 19:10:32 +02:00
|
|
|
|
|
|
|
|
|
2016-10-06 19:02:05 -07:00
|
|
|
|
|
|
|
|
|
2013-12-20 17:39:59 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-08-14 11:43:56 -07:00
|
|
|
|
2013-12-20 17:39:59 +08:00
|
|
|
|
2016-10-06 19:02:05 -07:00
|
|
|
|
2013-12-20 17:39:59 +08:00
|
|
|
|
2015-05-25 18:03:38 +08:00
|
|
|
|
2013-12-20 17:39:59 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-07-27 10:36:31 -07:00
|
|
|
|
|
|
|
|
|
2016-05-17 16:23:36 -07:00
|
|
|
|
2016-07-27 10:36:31 -07:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
|
|
|
|
|
2016-10-28 17:45:00 +09:00
|
|
|
|
2016-06-13 09:47:48 -07:00
|
|
|
|
2016-06-04 14:25:24 -07:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-06-04 14:25:24 -07:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
2016-12-21 12:13:03 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-06-05 14:31:41 -05:00
|
|
|
|
2016-05-17 16:23:36 -07:00
|
|
|
|
|
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-06-05 14:31:55 -05:00
|
|
|
|
|
|
|
|
|
2016-12-21 12:13:03 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-07-27 10:36:31 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
|
|
|
|
|
2016-01-18 18:24:59 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2016-01-18 18:24:59 +08:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2016-01-18 18:24:59 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-05-15 16:26:10 -07:00
|
|
|
|
2016-01-18 18:24:59 +08:00
|
|
|
|
2015-05-15 16:26:10 -07:00
|
|
|
|
|
|
|
|
|
2016-01-18 18:24:59 +08:00
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-18 18:24:59 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2014-03-22 14:57:23 +08:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-06-05 14:31:55 -05:00
|
|
|
|
2017-05-02 17:03:47 +02:00
|
|
|
|
2016-11-01 07:40:10 -06:00
|
|
|
|
2017-02-06 13:57:58 -08:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
|
|
|
|
|
2014-03-22 14:57:23 +08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2016-02-24 17:17:55 +08:00
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2016-02-24 17:17:55 +08:00
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-02-24 17:17:55 +08:00
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2017-04-05 22:49:44 +03:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2015-05-15 16:26:10 -07:00
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2014-12-24 16:08:14 +08:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2016-06-05 14:31:55 -05:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-06-05 14:31:55 -05:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2016-07-27 10:36:31 -07:00
|
|
|
|
2017-03-30 21:02:46 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
2017-05-19 23:37:01 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
f2fs: trace old block address for CoWed page
This patch enables to trace old block address of CoWed page for better
debugging.
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f0, oldaddr = 0xfe8ab, newaddr = 0xfee90 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f8, oldaddr = 0xfe8b0, newaddr = 0xfee91 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4fa, oldaddr = 0xfe8ae, newaddr = 0xfee92 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x96, oldaddr = 0xf049b, newaddr = 0x2bbe rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x97, oldaddr = 0xf049c, newaddr = 0x2bbf rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x98, oldaddr = 0xf049d, newaddr = 0x2bc0 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x47, oldaddr = 0xffffffff, newaddr = 0xf2631 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x48, oldaddr = 0xffffffff, newaddr = 0xf2632 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x49, oldaddr = 0xffffffff, newaddr = 0xf2633 rw = WRITE, type = DATA
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:36:38 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
f2fs: don't wait writeback for datas during checkpoint
Normally, while committing checkpoint, we will wait on all pages to be
writebacked no matter the page is data or metadata, so in scenario where
there are lots of data IO being submitted with metadata, we may suffer
long latency for waiting writeback during checkpoint.
Indeed, we only care about persistence for pages with metadata, but not
pages with data, as file system consistent are only related to metadate,
so in order to avoid encountering long latency in above scenario, let's
recognize and reference metadata in submitted IOs, wait writeback only
for metadatas.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-11-16 10:41:20 +08:00
|
|
|
|
|
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
f2fs: trace old block address for CoWed page
This patch enables to trace old block address of CoWed page for better
debugging.
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f0, oldaddr = 0xfe8ab, newaddr = 0xfee90 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f8, oldaddr = 0xfe8b0, newaddr = 0xfee91 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4fa, oldaddr = 0xfe8ae, newaddr = 0xfee92 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x96, oldaddr = 0xf049b, newaddr = 0x2bbe rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x97, oldaddr = 0xf049c, newaddr = 0x2bbf rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x98, oldaddr = 0xf049d, newaddr = 0x2bc0 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x47, oldaddr = 0xffffffff, newaddr = 0xf2631 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x48, oldaddr = 0xffffffff, newaddr = 0xf2632 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x49, oldaddr = 0xffffffff, newaddr = 0xf2633 rw = WRITE, type = DATA
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:36:38 +08:00
|
|
|
|
2016-10-06 19:02:05 -07:00
|
|
|
|
|
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
|
|
|
|
|
f2fs: trace old block address for CoWed page
This patch enables to trace old block address of CoWed page for better
debugging.
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f0, oldaddr = 0xfe8ab, newaddr = 0xfee90 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f8, oldaddr = 0xfe8b0, newaddr = 0xfee91 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4fa, oldaddr = 0xfe8ae, newaddr = 0xfee92 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x96, oldaddr = 0xf049b, newaddr = 0x2bbe rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x97, oldaddr = 0xf049c, newaddr = 0x2bbf rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x98, oldaddr = 0xf049d, newaddr = 0x2bc0 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x47, oldaddr = 0xffffffff, newaddr = 0xf2631 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x48, oldaddr = 0xffffffff, newaddr = 0xf2632 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x49, oldaddr = 0xffffffff, newaddr = 0xf2633 rw = WRITE, type = DATA
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:36:38 +08:00
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:18:25 -07:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: trace old block address for CoWed page
This patch enables to trace old block address of CoWed page for better
debugging.
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f0, oldaddr = 0xfe8ab, newaddr = 0xfee90 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f8, oldaddr = 0xfe8b0, newaddr = 0xfee91 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4fa, oldaddr = 0xfe8ae, newaddr = 0xfee92 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x96, oldaddr = 0xf049b, newaddr = 0x2bbe rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x97, oldaddr = 0xf049c, newaddr = 0x2bbf rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x98, oldaddr = 0xf049d, newaddr = 0x2bc0 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x47, oldaddr = 0xffffffff, newaddr = 0xf2631 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x48, oldaddr = 0xffffffff, newaddr = 0xf2632 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x49, oldaddr = 0xffffffff, newaddr = 0xf2633 rw = WRITE, type = DATA
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:36:38 +08:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2017-05-19 23:37:01 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
2014-03-22 14:57:23 +08:00
|
|
|
|
2016-12-14 10:12:56 -08:00
|
|
|
|
2013-11-30 12:51:14 +09:00
|
|
|
|
|
|
|
|
|
2017-09-06 21:04:44 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-19 00:19:06 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-19 00:19:06 +08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
2012-11-29 13:28:09 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-19 19:23:32 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-07 13:23:12 -08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2016-02-24 17:16:47 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2014-09-02 15:31:18 -07:00
|
|
|
|
2017-07-09 00:13:07 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-20 10:13:22 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2017-07-09 00:13:07 +08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-19 00:19:06 +08:00
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-11-10 23:13:18 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-27 17:04:17 +09:00
|
|
|
|
2013-11-10 23:13:18 +08:00
|
|
|
|
|
|
|
|
|
2013-12-27 17:04:17 +09:00
|
|
|
|
2013-11-10 23:13:18 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-08-05 15:52:16 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2017-02-23 09:18:05 +00:00
|
|
|
|
2015-08-05 15:52:16 +08:00
|
|
|
|
2015-03-19 19:26:02 +08:00
|
|
|
|
2015-08-05 15:52:16 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-02-05 17:54:31 +08:00
|
|
|
|
|
|
|
|
|
2015-08-05 15:52:16 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2015-10-09 15:11:38 -07:00
|
|
|
|
2016-06-05 14:31:55 -05:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-23 09:18:05 +00:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2015-10-09 15:11:38 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-02-05 18:03:40 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-05-13 08:38:35 +09:00
|
|
|
|
|
|
|
|
|
2015-07-15 13:08:21 -07:00
|
|
|
|
|
|
|
|
|
2013-05-13 08:38:35 +09:00
|
|
|
|
|
|
|
|
|
2013-12-06 15:00:58 +09:00
|
|
|
|
2015-07-15 13:08:21 -07:00
|
|
|
|
|
|
|
|
|
2013-05-13 08:38:35 +09:00
|
|
|
|
2015-02-05 18:03:40 +08:00
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-08-20 19:13:07 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2016-06-30 18:49:15 -07:00
|
|
|
|
|
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
2013-08-20 19:13:07 +09:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2017-09-06 21:04:44 -07:00
|
|
|
|
2013-03-08 21:29:23 +09:00
|
|
|
|
2015-07-15 13:08:21 -07:00
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
2015-07-15 13:08:21 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
2013-03-08 21:29:23 +09:00
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-01 07:40:10 -06:00
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-09 15:11:38 -07:00
|
|
|
|
|
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-01 07:40:10 -06:00
|
|
|
|
2015-04-30 17:00:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-03-08 21:29:23 +09:00
|
|
|
|
2013-12-06 15:00:58 +09:00
|
|
|
|
2013-04-26 11:55:17 +09:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-07-03 22:05:12 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-29 13:28:09 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
f2fs: introduce a new global lock scheme
In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.
Reference the following lock types in f2fs.h.
enum lock_type {
RENAME, /* for renaming operations */
DENTRY_OPS, /* for directory operations */
DATA_WRITE, /* for data write */
DATA_NEW, /* for data allocation */
DATA_TRUNC, /* for data truncate */
NODE_NEW, /* for node allocation */
NODE_TRUNC, /* for node truncate */
NODE_WRITE, /* for node write */
NR_LOCK_TYPE,
};
In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.
In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.
For this, I propose a new global lock scheme as follows.
0. Data structure
- f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
- f2fs_sb_info -> node_write
1. mutex_lock_op(sbi)
- try to get an avaiable lock from the array.
- returns the index of the gottern lock variable.
2. mutex_unlock_op(sbi, index of the lock)
- unlock the given index of the lock.
3. mutex_lock_all(sbi)
- grab all the locks in the array before the checkpoint.
4. mutex_unlock_all(sbi)
- release all the locks in the array after checkpoint.
5. block_operations()
- call mutex_lock_all()
- sync_dirty_dir_inodes()
- grab node_write
- sync_node_pages()
Note that,
the pairs of mutex_lock_op()/mutex_unlock_op() and
mutex_lock_all()/mutex_unlock_all() should be used together.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 16:21:29 +09:00
|
|
|
|
2013-12-21 18:02:14 +08:00
|
|
|
|
|
|
|
|
|
2015-07-14 18:14:06 +08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-05-20 09:55:50 +09:00
|
|
|
|
2013-12-27 17:04:17 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-01 22:03:47 -08:00
|
|
|
|
2015-10-09 15:11:38 -07:00
|
|
|
|
2015-07-14 18:14:06 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-29 11:18:42 -07:00
|
|
|
|
2015-07-14 18:14:06 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-12-27 17:04:17 +09:00
|
|
|
|
2013-11-10 23:13:18 +08:00
|
|
|
|
2015-04-29 11:18:42 -07:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-12-27 17:04:17 +09:00
|
|
|
|
2015-04-29 11:18:42 -07:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2015-04-29 11:18:42 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2016-06-30 18:49:15 -07:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2013-12-27 17:04:17 +09:00
|
|
|
|
2016-01-01 22:03:47 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2016-01-01 22:03:47 -08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-04-29 11:18:42 -07:00
|
|
|
|
2015-09-11 14:43:52 +08:00
|
|
|
|
2016-05-20 16:32:49 -07:00
|
|
|
|
2016-05-20 09:22:03 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
|
|
|
|
|
2014-09-02 15:31:18 -07:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
|
|
|
|
|
2014-09-15 19:32:16 -07:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
2017-07-09 00:13:07 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2016-05-20 10:13:22 -07:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
f2fs: preallocate fallocated blocks for direct IO
Normally, due to DIO_SKIP_HOLES flag is set by default, blockdev_direct_IO in
f2fs_direct_IO tries to skip DIO in holes when writing inside i_size, this
makes us falling back to buffered IO which shows lower performance.
So in commit 59b802e5a453 ("f2fs: allocate data blocks in advance for
f2fs_direct_IO"), we improve perfromance by allocating data blocks in advance
if we meet holes no matter in i_size or not, since with it we can avoid falling
back to buffered IO.
But we forget to consider for unwritten fallocated block in this commit.
This patch tries to fix it for fallocate case, this helps to improve
performance.
Test result:
Storage info: sandisk ultra 64G micro sd card.
touch /mnt/f2fs/file
truncate -s 67108864 /mnt/f2fs/file
fallocate -o 0 -l 67108864 /mnt/f2fs/file
time dd if=/dev/zero of=/mnt/f2fs/file bs=1M count=64 conv=notrunc oflag=direct
Time before applying the patch:
67108864 bytes (67 MB) copied, 36.16 s, 1.9 MB/s
real 0m36.162s
user 0m0.000s
sys 0m0.180s
Time after applying the patch:
67108864 bytes (67 MB) copied, 27.7776 s, 2.4 MB/s
real 0m27.780s
user 0m0.000s
sys 0m0.036s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-03-23 10:33:37 +08:00
|
|
|
|
2017-07-19 00:19:06 +08:00
|
|
|
|
|
|
|
|
|
f2fs: preallocate fallocated blocks for direct IO
Normally, due to DIO_SKIP_HOLES flag is set by default, blockdev_direct_IO in
f2fs_direct_IO tries to skip DIO in holes when writing inside i_size, this
makes us falling back to buffered IO which shows lower performance.
So in commit 59b802e5a453 ("f2fs: allocate data blocks in advance for
f2fs_direct_IO"), we improve perfromance by allocating data blocks in advance
if we meet holes no matter in i_size or not, since with it we can avoid falling
back to buffered IO.
But we forget to consider for unwritten fallocated block in this commit.
This patch tries to fix it for fallocate case, this helps to improve
performance.
Test result:
Storage info: sandisk ultra 64G micro sd card.
touch /mnt/f2fs/file
truncate -s 67108864 /mnt/f2fs/file
fallocate -o 0 -l 67108864 /mnt/f2fs/file
time dd if=/dev/zero of=/mnt/f2fs/file bs=1M count=64 conv=notrunc oflag=direct
Time before applying the patch:
67108864 bytes (67 MB) copied, 36.16 s, 1.9 MB/s
real 0m36.162s
user 0m0.000s
sys 0m0.180s
Time after applying the patch:
67108864 bytes (67 MB) copied, 27.7776 s, 2.4 MB/s
real 0m27.780s
user 0m0.000s
sys 0m0.036s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-03-23 10:33:37 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-09 00:13:07 +08:00
|
|
|
|
|
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
f2fs: preallocate fallocated blocks for direct IO
Normally, due to DIO_SKIP_HOLES flag is set by default, blockdev_direct_IO in
f2fs_direct_IO tries to skip DIO in holes when writing inside i_size, this
makes us falling back to buffered IO which shows lower performance.
So in commit 59b802e5a453 ("f2fs: allocate data blocks in advance for
f2fs_direct_IO"), we improve perfromance by allocating data blocks in advance
if we meet holes no matter in i_size or not, since with it we can avoid falling
back to buffered IO.
But we forget to consider for unwritten fallocated block in this commit.
This patch tries to fix it for fallocate case, this helps to improve
performance.
Test result:
Storage info: sandisk ultra 64G micro sd card.
touch /mnt/f2fs/file
truncate -s 67108864 /mnt/f2fs/file
fallocate -o 0 -l 67108864 /mnt/f2fs/file
time dd if=/dev/zero of=/mnt/f2fs/file bs=1M count=64 conv=notrunc oflag=direct
Time before applying the patch:
67108864 bytes (67 MB) copied, 36.16 s, 1.9 MB/s
real 0m36.162s
user 0m0.000s
sys 0m0.180s
Time after applying the patch:
67108864 bytes (67 MB) copied, 27.7776 s, 2.4 MB/s
real 0m27.780s
user 0m0.000s
sys 0m0.036s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-03-23 10:33:37 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: preallocate fallocated blocks for direct IO
Normally, due to DIO_SKIP_HOLES flag is set by default, blockdev_direct_IO in
f2fs_direct_IO tries to skip DIO in holes when writing inside i_size, this
makes us falling back to buffered IO which shows lower performance.
So in commit 59b802e5a453 ("f2fs: allocate data blocks in advance for
f2fs_direct_IO"), we improve perfromance by allocating data blocks in advance
if we meet holes no matter in i_size or not, since with it we can avoid falling
back to buffered IO.
But we forget to consider for unwritten fallocated block in this commit.
This patch tries to fix it for fallocate case, this helps to improve
performance.
Test result:
Storage info: sandisk ultra 64G micro sd card.
touch /mnt/f2fs/file
truncate -s 67108864 /mnt/f2fs/file
fallocate -o 0 -l 67108864 /mnt/f2fs/file
time dd if=/dev/zero of=/mnt/f2fs/file bs=1M count=64 conv=notrunc oflag=direct
Time before applying the patch:
67108864 bytes (67 MB) copied, 36.16 s, 1.9 MB/s
real 0m36.162s
user 0m0.000s
sys 0m0.180s
Time after applying the patch:
67108864 bytes (67 MB) copied, 27.7776 s, 2.4 MB/s
real 0m27.780s
user 0m0.000s
sys 0m0.036s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-03-23 10:33:37 +08:00
|
|
|
|
2017-05-19 23:37:01 +08:00
|
|
|
|
2015-03-19 19:23:32 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2014-09-15 19:32:16 -07:00
|
|
|
|
2016-01-26 15:39:35 +08:00
|
|
|
|
2014-09-15 19:32:16 -07:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2016-05-20 09:22:03 -07:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-11 16:46:40 -08:00
|
|
|
|
|
|
|
|
|
2017-09-05 16:54:24 -07:00
|
|
|
|
2016-11-11 16:46:40 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
2015-02-09 12:09:53 -08:00
|
|
|
|
2016-02-03 13:09:09 -08:00
|
|
|
|
2016-01-26 15:38:29 +08:00
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
2015-02-09 12:09:53 -08:00
|
|
|
|
2017-01-13 13:12:29 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-07 08:52:57 -07:00
|
|
|
|
2016-08-20 15:12:01 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2015-02-09 12:09:53 -08:00
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-11 16:46:40 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-02-03 13:09:09 -08:00
|
|
|
|
2017-07-19 00:19:05 +08:00
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
2015-02-09 12:09:53 -08:00
|
|
|
|
|
|
|
|
|
2017-03-13 20:22:18 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-29 13:28:09 +09:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
|
|
|
|
|
2013-12-21 18:02:14 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-10-27 09:53:45 +08:00
|
|
|
|
2015-08-19 19:11:19 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-09-21 20:17:52 +08:00
|
|
|
|
2016-06-16 17:03:23 -07:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
2017-02-23 09:18:05 +00:00
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-08-20 15:12:01 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2015-02-05 17:50:30 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
2015-02-09 12:09:53 -08:00
|
|
|
|
2017-03-13 20:22:18 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2013-12-26 16:55:22 +09:00
|
|
|
|
2016-05-05 19:13:02 +08:00
|
|
|
|
|
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2013-04-23 16:38:02 +09:00
|
|
|
|
2015-09-18 16:51:51 +08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
2016-11-02 14:52:15 +01:00
|
|
|
|
2016-01-26 15:39:35 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
2017-07-19 00:19:06 +08:00
|
|
|
|
2015-09-18 16:51:51 +08:00
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-09-18 16:51:51 +08:00
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
|
|
|
|
|
2016-10-11 22:57:03 +08:00
|
|
|
|
2016-05-20 10:13:22 -07:00
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
|
|
|
|
|
2017-02-23 19:55:05 +08:00
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
|
|
|
|
|
2016-05-05 19:13:02 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
2016-05-05 19:13:02 +08:00
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
|
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
2016-02-03 13:09:09 -08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
|
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:37:38 +08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
2015-12-17 13:20:59 +08:00
|
|
|
|
|
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-03-13 20:22:18 +08:00
|
|
|
|
2016-10-11 22:57:03 +08:00
|
|
|
|
2016-05-09 19:56:30 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2015-12-22 13:23:35 -08:00
|
|
|
|
2017-03-13 20:22:18 +08:00
|
|
|
|
2016-10-11 22:57:03 +08:00
|
|
|
|
2015-12-22 13:23:35 -08:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
2013-12-16 19:04:05 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
|
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
|
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
|
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
|
|
|
|
|
2017-01-22 12:21:02 +08:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
2016-11-11 16:31:56 -08:00
|
|
|
|
2015-04-06 19:55:34 -07:00
|
|
|
|
|
|
|
|
|
2014-06-13 13:02:11 +09:00
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
|
|
|
|
|
2014-06-13 13:02:11 +09:00
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
|
|
|
|
|
2014-06-13 13:02:11 +09:00
|
|
|
|
|
|
|
|
|
2015-08-19 19:11:19 +08:00
|
|
|
|
2014-06-13 13:02:11 +09:00
|
|
|
|
|
|
|
|
|
2015-08-19 19:11:19 +08:00
|
|
|
|
2017-08-09 17:27:30 +08:00
|
|
|
|
2015-08-19 19:11:19 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-28 21:48:32 +08:00
|
|
|
|
2015-12-31 14:35:37 +08:00
|
|
|
|
2015-12-28 21:48:32 +08:00
|
|
|
|
|
|
|
|
|
2015-08-19 19:11:19 +08:00
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2014-06-13 13:02:11 +09:00
|
|
|
|
|
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-06-08 04:30:14 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
|
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-15 11:34:49 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-22 15:40:57 -05:00
|
|
|
|
2016-01-04 15:56:50 +08:00
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-26 18:07:41 +08:00
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-08-19 19:11:19 +08:00
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2016-10-11 22:57:04 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-26 18:07:41 +08:00
|
|
|
|
2016-10-11 22:57:04 +08:00
|
|
|
|
2015-12-26 18:07:41 +08:00
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
2015-12-26 18:07:41 +08:00
|
|
|
|
2016-01-08 20:19:27 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-26 18:07:41 +08:00
|
|
|
|
|
|
|
|
|
2016-01-08 20:19:27 +08:00
|
|
|
|
2015-12-26 18:07:41 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-22 15:40:57 -05:00
|
|
|
|
2015-05-08 19:30:32 -07:00
|
|
|
|
2014-06-08 04:30:14 +09:00
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-26 15:42:58 +08:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-01-07 18:49:42 +08:00
|
|
|
|
2017-03-13 16:35:13 +08:00
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
2016-07-26 15:24:53 -07:00
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-09-18 16:54:16 +08:00
|
|
|
|
2017-08-09 17:27:30 +08:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2016-06-30 18:49:15 -07:00
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-10-06 19:02:05 -07:00
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
2016-07-27 10:36:31 -07:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-09-06 21:04:44 -07:00
|
|
|
|
2016-07-12 09:38:48 -07:00
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-07-27 10:36:31 -07:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-07-27 10:36:31 -07:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
2014-10-23 19:48:09 -07:00
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
2014-05-06 16:53:08 +08:00
|
|
|
|
|
|
|
|
|
2014-08-06 23:22:50 +09:00
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
|
|
|
|
|
2014-10-23 19:48:09 -07:00
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
2017-01-07 18:49:42 +08:00
|
|
|
|
2015-10-12 17:02:26 +08:00
|
|
|
|
|
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-09 11:20:42 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2017-04-25 12:45:12 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-09-05 16:54:24 -07:00
|
|
|
|
2017-04-25 12:45:12 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-09-05 17:04:35 -07:00
|
|
|
|
2017-04-25 12:45:12 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2017-04-25 12:45:12 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-24 15:20:16 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2017-04-25 12:45:13 +00:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-25 12:45:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-24 15:20:16 -07:00
|
|
|
|
|
|
|
|
|
2017-04-25 12:45:13 +00:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
2017-04-25 12:45:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2017-06-21 17:52:39 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2013-02-26 13:10:46 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
f2fs: support revoking atomic written pages
f2fs support atomic write with following semantics:
1. open db file
2. ioctl start atomic write
3. (write db file) * n
4. ioctl commit atomic write
5. close db file
With this flow we can avoid file becoming corrupted when abnormal power
cut, because we hold data of transaction in referenced pages linked in
inmem_pages list of inode, but without setting them dirty, so these data
won't be persisted unless we commit them in step 4.
But we should still hold journal db file in memory by using volatile
write, because our semantics of 'atomic write support' is incomplete, in
step 4, we could fail to submit all dirty data of transaction, once
partial dirty data was committed in storage, then after a checkpoint &
abnormal power-cut, db file will be corrupted forever.
So this patch tries to improve atomic write flow by adding a revoking flow,
once inner error occurs in committing, this gives another chance to try to
revoke these partial submitted data of current transaction, it makes
committing operation more like aotmical one.
If we're not lucky, once revoking operation was failed, EAGAIN will be
reported to user for suggesting doing the recovery with held journal file,
or retrying current transaction again.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-06 14:40:34 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
f2fs: trace old block address for CoWed page
This patch enables to trace old block address of CoWed page for better
debugging.
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f0, oldaddr = 0xfe8ab, newaddr = 0xfee90 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4f8, oldaddr = 0xfe8b0, newaddr = 0xfee91 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 1, page_index = 0x1d4fa, oldaddr = 0xfe8ae, newaddr = 0xfee92 rw = WRITE_SYNC, type = NODE
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x96, oldaddr = 0xf049b, newaddr = 0x2bbe rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x97, oldaddr = 0xf049c, newaddr = 0x2bbf rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 134824, page_index = 0x98, oldaddr = 0xf049d, newaddr = 0x2bc0 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x47, oldaddr = 0xffffffff, newaddr = 0xf2631 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x48, oldaddr = 0xffffffff, newaddr = 0xf2632 rw = WRITE, type = DATA
f2fs_submit_page_mbio: dev = (1,0), ino = 135260, page_index = 0x49, oldaddr = 0xffffffff, newaddr = 0xf2633 rw = WRITE, type = DATA
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:36:38 +08:00
|
|
|
|
2015-02-25 19:25:01 -08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-02-25 19:25:01 -08:00
|
|
|
|
2017-04-25 12:45:13 +00:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-24 15:20:16 -07:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2017-03-30 21:02:46 -07:00
|
|
|
|
2017-04-25 12:45:12 +00:00
|
|
|
|
2016-05-20 10:13:22 -07:00
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2014-09-02 15:31:18 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2016-05-20 20:42:37 -07:00
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
f2fs: introduce a new global lock scheme
In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.
Reference the following lock types in f2fs.h.
enum lock_type {
RENAME, /* for renaming operations */
DENTRY_OPS, /* for directory operations */
DATA_WRITE, /* for data write */
DATA_NEW, /* for data allocation */
DATA_TRUNC, /* for data truncate */
NODE_NEW, /* for node allocation */
NODE_TRUNC, /* for node truncate */
NODE_WRITE, /* for node write */
NR_LOCK_TYPE,
};
In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.
In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.
For this, I propose a new global lock scheme as follows.
0. Data structure
- f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
- f2fs_sb_info -> node_write
1. mutex_lock_op(sbi)
- try to get an avaiable lock from the array.
- returns the index of the gottern lock variable.
2. mutex_unlock_op(sbi, index of the lock)
- unlock the given index of the lock.
3. mutex_lock_all(sbi)
- grab all the locks in the array before the checkpoint.
4. mutex_unlock_all(sbi)
- release all the locks in the array after checkpoint.
5. block_operations()
- call mutex_lock_all()
- sync_dirty_dir_inodes()
- grab node_write
- sync_node_pages()
Note that,
the pairs of mutex_lock_op()/mutex_unlock_op() and
mutex_lock_all()/mutex_unlock_all() should be used together.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 16:21:29 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2016-06-05 14:31:55 -05:00
|
|
|
|
2016-11-01 10:00:38 -06:00
|
|
|
|
2017-04-25 12:45:13 +00:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2014-05-06 16:48:26 +08:00
|
|
|
|
|
|
|
|
|
2017-06-29 23:20:45 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
f2fs: introduce a new global lock scheme
In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.
Reference the following lock types in f2fs.h.
enum lock_type {
RENAME, /* for renaming operations */
DENTRY_OPS, /* for directory operations */
DATA_WRITE, /* for data write */
DATA_NEW, /* for data allocation */
DATA_TRUNC, /* for data truncate */
NODE_NEW, /* for node allocation */
NODE_TRUNC, /* for node truncate */
NODE_WRITE, /* for node write */
NR_LOCK_TYPE,
};
In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.
In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.
For this, I propose a new global lock scheme as follows.
0. Data structure
- f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
- f2fs_sb_info -> node_write
1. mutex_lock_op(sbi)
- try to get an avaiable lock from the array.
- returns the index of the gottern lock variable.
2. mutex_unlock_op(sbi, index of the lock)
- unlock the given index of the lock.
3. mutex_lock_all(sbi)
- grab all the locks in the array before the checkpoint.
4. mutex_unlock_all(sbi)
- release all the locks in the array after checkpoint.
5. block_operations()
- call mutex_lock_all()
- sync_dirty_dir_inodes()
- grab node_write
- sync_node_pages()
Note that,
the pairs of mutex_lock_op()/mutex_unlock_op() and
mutex_lock_all()/mutex_unlock_all() should be used together.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 16:21:29 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2014-04-15 16:04:15 +09:00
|
|
|
|
f2fs: introduce a new global lock scheme
In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.
Reference the following lock types in f2fs.h.
enum lock_type {
RENAME, /* for renaming operations */
DENTRY_OPS, /* for directory operations */
DATA_WRITE, /* for data write */
DATA_NEW, /* for data allocation */
DATA_TRUNC, /* for data truncate */
NODE_NEW, /* for node allocation */
NODE_TRUNC, /* for node truncate */
NODE_WRITE, /* for node write */
NR_LOCK_TYPE,
};
In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.
In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.
For this, I propose a new global lock scheme as follows.
0. Data structure
- f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
- f2fs_sb_info -> node_write
1. mutex_lock_op(sbi)
- try to get an avaiable lock from the array.
- returns the index of the gottern lock variable.
2. mutex_unlock_op(sbi, index of the lock)
- unlock the given index of the lock.
3. mutex_lock_all(sbi)
- grab all the locks in the array before the checkpoint.
4. mutex_unlock_all(sbi)
- release all the locks in the array after checkpoint.
5. block_operations()
- call mutex_lock_all()
- sync_dirty_dir_inodes()
- grab node_write
- sync_node_pages()
Note that,
the pairs of mutex_lock_op()/mutex_unlock_op() and
mutex_lock_all()/mutex_unlock_all() should be used together.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 16:21:29 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
f2fs: introduce a new global lock scheme
In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.
Reference the following lock types in f2fs.h.
enum lock_type {
RENAME, /* for renaming operations */
DENTRY_OPS, /* for directory operations */
DATA_WRITE, /* for data write */
DATA_NEW, /* for data allocation */
DATA_TRUNC, /* for data truncate */
NODE_NEW, /* for node allocation */
NODE_TRUNC, /* for node truncate */
NODE_WRITE, /* for node write */
NR_LOCK_TYPE,
};
In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.
In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.
For this, I propose a new global lock scheme as follows.
0. Data structure
- f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
- f2fs_sb_info -> node_write
1. mutex_lock_op(sbi)
- try to get an avaiable lock from the array.
- returns the index of the gottern lock variable.
2. mutex_unlock_op(sbi, index of the lock)
- unlock the given index of the lock.
3. mutex_lock_all(sbi)
- grab all the locks in the array before the checkpoint.
4. mutex_unlock_all(sbi)
- release all the locks in the array after checkpoint.
5. block_operations()
- call mutex_lock_all()
- sync_dirty_dir_inodes()
- grab node_write
- sync_node_pages()
Note that,
the pairs of mutex_lock_op()/mutex_unlock_op() and
mutex_lock_all()/mutex_unlock_all() should be used together.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 16:21:29 +09:00
|
|
|
|
2014-12-09 06:08:59 -08:00
|
|
|
|
|
|
|
|
|
2016-04-14 16:48:52 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-12-09 06:08:59 -08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2014-08-11 18:37:46 -07:00
|
|
|
|
|
|
|
|
|
2016-06-03 12:28:26 -07:00
|
|
|
|
2014-09-12 15:53:45 -07:00
|
|
|
|
2014-08-11 18:37:46 -07:00
|
|
|
|
|
|
|
|
|
2016-05-29 21:18:23 -07:00
|
|
|
|
|
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
2016-05-29 21:18:23 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-02-17 19:29:27 +09:00
|
|
|
|
|
|
|
|
|
2016-09-01 12:02:51 -07:00
|
|
|
|
f2fs: introduce a new global lock scheme
In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.
Reference the following lock types in f2fs.h.
enum lock_type {
RENAME, /* for renaming operations */
DENTRY_OPS, /* for directory operations */
DATA_WRITE, /* for data write */
DATA_NEW, /* for data allocation */
DATA_TRUNC, /* for data truncate */
NODE_NEW, /* for node allocation */
NODE_TRUNC, /* for node truncate */
NODE_WRITE, /* for node write */
NR_LOCK_TYPE,
};
In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.
In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.
For this, I propose a new global lock scheme as follows.
0. Data structure
- f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
- f2fs_sb_info -> node_write
1. mutex_lock_op(sbi)
- try to get an avaiable lock from the array.
- returns the index of the gottern lock variable.
2. mutex_unlock_op(sbi, index of the lock)
- unlock the given index of the lock.
3. mutex_lock_all(sbi)
- grab all the locks in the array before the checkpoint.
4. mutex_unlock_all(sbi)
- release all the locks in the array after checkpoint.
5. block_operations()
- call mutex_lock_all()
- sync_dirty_dir_inodes()
- grab node_write
- sync_node_pages()
Note that,
the pairs of mutex_lock_op()/mutex_unlock_op() and
mutex_lock_all()/mutex_unlock_all() should be used together.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 16:21:29 +09:00
|
|
|
|
2017-03-24 20:05:13 -04:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2014-10-23 19:48:09 -07:00
|
|
|
|
2017-02-23 20:31:20 +08:00
|
|
|
|
2014-10-23 19:48:09 -07:00
|
|
|
|
2017-02-23 20:31:20 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
2015-04-23 14:38:15 -07:00
|
|
|
|
2017-05-12 13:51:34 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-20 20:42:37 -07:00
|
|
|
|
|
|
|
|
|
2017-04-27 00:17:21 +08:00
|
|
|
|
2014-02-17 19:29:27 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
f2fs: introduce a new global lock scheme
In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.
Reference the following lock types in f2fs.h.
enum lock_type {
RENAME, /* for renaming operations */
DENTRY_OPS, /* for directory operations */
DATA_WRITE, /* for data write */
DATA_NEW, /* for data allocation */
DATA_TRUNC, /* for data truncate */
NODE_NEW, /* for node allocation */
NODE_TRUNC, /* for node truncate */
NODE_WRITE, /* for node write */
NR_LOCK_TYPE,
};
In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.
In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.
For this, I propose a new global lock scheme as follows.
0. Data structure
- f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
- f2fs_sb_info -> node_write
1. mutex_lock_op(sbi)
- try to get an avaiable lock from the array.
- returns the index of the gottern lock variable.
2. mutex_unlock_op(sbi, index of the lock)
- unlock the given index of the lock.
3. mutex_lock_all(sbi)
- grab all the locks in the array before the checkpoint.
4. mutex_unlock_all(sbi)
- release all the locks in the array after checkpoint.
5. block_operations()
- call mutex_lock_all()
- sync_dirty_dir_inodes()
- grab node_write
- sync_node_pages()
Note that,
the pairs of mutex_lock_op()/mutex_unlock_op() and
mutex_lock_all()/mutex_unlock_all() should be used together.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 16:21:29 +09:00
|
|
|
|
2014-09-12 15:53:45 -07:00
|
|
|
|
2015-02-25 19:25:01 -08:00
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2017-03-24 20:05:13 -04:00
|
|
|
|
2015-12-16 13:09:20 +08:00
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
2015-11-10 18:45:07 +08:00
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
|
|
|
|
|
2017-04-20 13:51:57 -07:00
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: introduce f2fs_submit_merged_bio_cond
f2fs use single bio buffer per type data (META/NODE/DATA) for caching
writes locating in continuous block address as many as possible, after
submitting, these writes may be still cached in bio buffer, so we have
to flush cached writes in bio buffer by calling f2fs_submit_merged_bio.
Unfortunately, in the scenario of high concurrency, bio buffer could be
flushed by someone else before we submit it as below reasons:
a) there is no space in bio buffer.
b) add a request of different type (SYNC, ASYNC).
c) add a discontinuous block address.
For this condition, f2fs_submit_merged_bio will be devastating, because
it could break the following merging of writes in bio buffer, split one
big bio into two smaller one.
This patch introduces f2fs_submit_merged_bio_cond which can do a
conditional submitting with bio buffer, before submitting it will judge
whether:
- page in DATA type bio buffer is matching with specified page;
- page in DATA type bio buffer is belong to specified inode;
- page in NODE type bio buffer is belong to specified inode;
If there is no eligible page in bio buffer, we will skip submitting step,
result in gaining more chance to merge consecutive block IOs in bio cache.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-18 18:28:11 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-15 16:04:15 +09:00
|
|
|
|
2016-11-28 19:13:43 -08:00
|
|
|
|
|
|
|
|
|
2016-05-29 21:18:23 -07:00
|
|
|
|
|
|
|
|
|
2013-01-15 16:45:24 +09:00
|
|
|
|
|
|
|
|
|
2017-02-03 17:18:00 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
2017-02-03 17:18:00 -08:00
|
|
|
|
|
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
|
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-25 20:57:16 -07:00
|
|
|
|
2017-03-24 20:05:13 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
|
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-21 17:52:39 -07:00
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-20 23:43:51 +08:00
|
|
|
|
|
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
2016-11-28 19:13:43 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-21 17:52:39 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-28 19:13:43 -08:00
|
|
|
|
2016-05-29 21:18:23 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-03 17:44:04 -08:00
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
2017-03-28 18:07:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-01 16:51:22 -08:00
|
|
|
|
2017-05-10 11:28:38 -07:00
|
|
|
|
|
|
|
|
|
2016-09-29 18:50:11 +08:00
|
|
|
|
f2fs: expose f2fs_write_cache_pages
If there are gced dirty pages and normal dirty pages in the mapping
of one inode, we might writeback them alternately with discontinuous
block address, resulting in low performance.
This patch introduces f2fs_write_cache_pages with codes copied from
write_cache_pages in mm/page-writeback.c.
In this function, we refactor flow with two steps:
1) writeback all cold type pages.
2) writeback all non-cold type pages.
By using this method, f2fs will writeback dirty pages with the same
temperature in bunch mode, it makes writeouted block being with
more continuous address, so they can be merged as much as possible
in f2fs bio cache, and also it will reduce the chance of submiting
small IO from block layer.
Test environment: 8g nokia sd card (very old sd card, but it shows
better effect when testing with this patch, and with a 32g kingston
sd card, I didn't see much more improvement).
Test step:
1. touch testfile;
2. truncate -s 512K testfile;
3. write all pages with odd index;
4. trigger gc by ioctl;
5. write all pages with even index;
6. time fsync testfile.
before:
real 0m0.402s
user 0m0.000s
sys 0m0.000s
after:
real 0m0.143s
user 0m0.004s
sys 0m0.004s
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-14 18:56:10 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2014-09-02 15:31:18 -07:00
|
|
|
|
2016-07-13 19:33:19 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2013-04-03 11:38:00 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-07-17 18:02:39 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-29 23:20:45 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-08 10:40:07 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-27 09:53:45 +08:00
|
|
|
|
2016-05-20 10:13:22 -07:00
|
|
|
|
2015-10-27 09:53:45 +08:00
|
|
|
|
|
|
|
|
|
2016-02-04 16:14:00 +08:00
|
|
|
|
|
|
|
|
|
2017-03-28 18:07:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-07-13 19:33:19 -07:00
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
2016-07-13 19:33:19 -07:00
|
|
|
|
2017-03-28 18:07:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-25 17:17:56 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-11 13:54:01 +09:00
|
|
|
|
2015-12-16 13:09:20 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2014-03-18 13:43:05 +09:00
|
|
|
|
|
|
|
|
|
2014-09-12 15:53:45 -07:00
|
|
|
|
2016-02-04 16:14:00 +08:00
|
|
|
|
2014-03-18 13:43:05 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-28 13:48:11 -08:00
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
2015-12-28 13:48:11 -08:00
|
|
|
|
2017-05-18 11:06:45 +08:00
|
|
|
|
2015-12-28 13:48:11 -08:00
|
|
|
|
|
|
|
|
|
2017-05-18 11:06:45 +08:00
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
2017-02-23 09:18:05 +00:00
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-01-13 13:12:29 -08:00
|
|
|
|
|
|
|
|
|
2016-02-03 13:49:44 -08:00
|
|
|
|
|
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2017-03-13 20:22:18 +08:00
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-19 00:19:05 +08:00
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
2016-05-20 10:13:22 -07:00
|
|
|
|
2016-05-11 19:48:44 +08:00
|
|
|
|
|
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-06 11:27:03 -07:00
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
2017-03-13 20:22:18 +08:00
|
|
|
|
|
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
2015-12-23 13:48:58 -08:00
|
|
|
|
2017-03-13 20:22:18 +08:00
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-09-02 15:31:18 -07:00
|
|
|
|
2015-07-15 13:08:21 -07:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2014-05-06 16:46:04 +08:00
|
|
|
|
|
|
|
|
|
2014-11-25 10:59:45 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-26 11:55:17 +09:00
|
|
|
|
2017-02-17 09:55:55 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-30 09:22:45 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
|
|
|
|
|
2014-10-17 20:33:55 -07:00
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
2014-10-17 20:33:55 -07:00
|
|
|
|
2016-09-01 12:02:51 -07:00
|
|
|
|
2015-12-22 13:23:35 -08:00
|
|
|
|
2016-01-07 14:15:04 -08:00
|
|
|
|
2015-12-22 13:23:35 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-01-20 23:43:51 +08:00
|
|
|
|
2014-10-23 19:48:09 -07:00
|
|
|
|
2015-10-08 13:27:34 +08:00
|
|
|
|
2017-09-05 16:54:24 -07:00
|
|
|
|
2017-09-05 17:04:35 -07:00
|
|
|
|
2015-10-08 13:27:34 +08:00
|
|
|
|
2016-09-06 11:02:03 -07:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-12-20 11:11:35 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-23 11:55:18 -08:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2016-09-06 11:02:03 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2017-09-06 21:04:44 -07:00
|
|
|
|
|
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
2014-03-29 15:30:40 +08:00
|
|
|
|
2013-03-08 21:29:23 +09:00
|
|
|
|
2013-12-06 15:00:58 +09:00
|
|
|
|
2013-04-26 11:55:17 +09:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-07-03 22:05:12 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-23 12:04:33 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2014-10-17 20:33:55 -07:00
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
2015-07-15 13:08:21 -07:00
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2013-06-27 13:04:08 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-05-06 16:47:23 +08:00
|
|
|
|
|
|
|
|
|
2016-09-06 11:02:03 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-12-20 11:11:35 +08:00
|
|
|
|
2016-09-06 11:02:03 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-10-09 13:19:53 -07:00
|
|
|
|
2013-06-27 13:04:08 +09:00
|
|
|
|
2016-05-20 09:22:03 -07:00
|
|
|
|
|
|
|
|
|
2016-09-06 11:02:03 -07:00
|
|
|
|
2016-08-06 21:09:41 +08:00
|
|
|
|
2016-01-08 16:57:48 -08:00
|
|
|
|
2013-06-27 13:04:08 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-16 04:33:52 -07:00
|
|
|
|
|
|
|
|
|
2013-12-26 20:15:09 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-03-16 18:07:34 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-26 20:15:09 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-07 08:51:58 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-02-03 13:09:09 -08:00
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
|
|
|
|
|
2016-04-07 08:51:58 -07:00
|
|
|
|
2016-07-13 09:18:29 +08:00
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
2013-12-26 20:15:09 +09:00
|
|
|
|
2016-02-03 13:09:09 -08:00
|
|
|
|
2015-12-22 11:09:35 -08:00
|
|
|
|
|
|
|
|
|
f2fs: handle inline data operations
Hook inline data read/write, truncate, fallocate, setattr, etc.
Files need meet following 2 requirement to inline:
1) file size is not greater than MAX_INLINE_DATA;
2) file doesn't pre-allocate data blocks by fallocate().
FI_INLINE_DATA will not be set while creating a new regular inode because
most of the files are bigger than ~3.4K. Set FI_INLINE_DATA only when
data is submitted to block layer, ranther than set it while creating a new
inode, this also avoids converting data from inline to normal data block
and vice versa.
While writting inline data to inode block, the first data block should be
released if the file has a block indexed by i_addr[0].
On the other hand, when a file operation is appied to a file with inline
data, we need to test if this file can remain inline by doing this
operation, otherwise it should be convert into normal file by reserving
a new data block, copying inline data to this new block and clear
FI_INLINE_DATA flag. Because reserve a new data block here will make use
of i_addr[0], if we save inline data in i_addr[0..872], then the first
4 bytes would be overwriten. This problem can be avoided simply by
not using i_addr[0] for inline data.
Signed-off-by: Huajun Li <huajun.li@intel.com>
Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Weihong Xu <weihong.xu@intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-11-10 23:13:20 +08:00
|
|
|
|
2016-11-11 16:46:40 -08:00
|
|
|
|
2016-06-03 19:29:38 -07:00
|
|
|
|
2015-04-21 20:39:58 -07:00
|
|
|
|
2016-07-22 15:25:47 -07:00
|
|
|
|
2014-07-31 21:11:22 +08:00
|
|
|
|
2016-07-13 09:18:29 +08:00
|
|
|
|
2016-04-07 08:51:58 -07:00
|
|
|
|
2016-07-13 09:18:29 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-20 10:13:22 -07:00
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
2016-04-18 17:07:44 -04:00
|
|
|
|
2017-08-02 23:21:48 +08:00
|
|
|
|
2016-04-18 17:07:44 -04:00
|
|
|
|
2014-07-31 21:11:22 +08:00
|
|
|
|
2016-07-22 15:25:47 -07:00
|
|
|
|
2014-07-31 21:11:22 +08:00
|
|
|
|
2014-07-02 13:25:04 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
2014-09-12 15:53:45 -07:00
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
|
|
|
|
2014-09-12 15:53:45 -07:00
|
|
|
|
|
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
2016-10-11 22:57:01 +08:00
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
2016-10-11 22:57:01 +08:00
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
2016-10-11 22:57:01 +08:00
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
2016-10-11 22:57:01 +08:00
|
|
|
|
|
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
2015-08-07 18:42:09 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-03-17 09:55:52 +08:00
|
|
|
|
2015-08-07 18:42:09 +08:00
|
|
|
|
2016-04-29 20:09:15 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-01-30 11:39:08 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-08-07 18:42:09 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-29 20:09:15 +08:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-03-14 09:24:32 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
2016-06-30 18:40:10 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-10-24 17:53:29 +09:00
|
|
|
|
|
|
|
|
|
2016-06-30 18:49:15 -07:00
|
|
|
|
|
|
|
|
|
2014-10-09 13:19:53 -07:00
|
|
|
|
2017-01-07 18:50:26 +08:00
|
|
|
|
2015-08-07 18:42:09 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-10-09 13:19:53 -07:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2016-06-30 18:40:10 -07:00
|
|
|
|
2014-09-12 15:53:45 -07:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-01-17 20:30:23 +09:00
|
|
|
|
|
|
|
|
|
2014-04-22 13:34:01 +08:00
|
|
|
|
|
|
|
|
|
2015-10-19 10:29:51 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-08-19 19:11:19 +08:00
|
|
|
|
2015-02-05 17:54:31 +08:00
|
|
|
|
|
|
|
|
|
2016-09-20 05:03:27 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-06 14:46:01 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-09-20 05:03:27 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-09-08 16:12:06 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-09-20 05:03:27 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-06-27 13:04:08 +09:00
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2015-02-05 17:44:29 +08:00
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
2013-01-17 20:30:23 +09:00
|
|
|
|
2016-09-20 05:03:27 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-02 17:10:12 +09:00
|
|
|
|
