During test patch that adjust page_size_mask to map small range ram with
big page size, found page table is setup wrongly for 32bit. And
native_pagetable_init wrong clear pte for pmd with large page support.
1. add more comments about why we are expecting pte.
2. add BUG checking, so next time we could find problem earlier
when we mess up page table setup again.
3. max_low_pfn is not included boundary for low memory mapping.
We should check from max_low_pfn instead of +1.
4. add print out when some pte really get cleared, or we should use
WARN() to find out why above max_low_pfn get mapped? so we could
fix it.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-35-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Put it in mm/init.c, and call it from probe_page_mask().
init_mem_mapping is calling probe_page_mask at first.
So calling sequence is not changed.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-32-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
On 32bit, before patcheset that only set page table for ram, we only
call that one time.
Now, we are calling that during every init_memory_mapping if we have holes
under max_low_pfn.
We should only call it one time after all ranges under max_low_page get
mapped just like we did before.
Also that could avoid the risk to run out of pgt_buf in BRK.
Need to update page_table_range_init() to count the pages for kmap page table
at first, and use new added alloc_low_pages() to get pages in sequence.
That will conform to the requirement that pages need to be in low to high order.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-30-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
32bit kmap mapping needs pages to be used for low to high.
At this point those pages are still from pgt_buf_* from BRK, so it is
ok now.
But we want to move early_ioremap_page_table_range_init() out of
init_memory_mapping() and only call it one time later, that will
make page_table_range_init/page_table_kmap_check/alloc_low_page to
use memblock to get page.
memblock allocation for pages are from high to low.
So will get panic from page_table_kmap_check() that has BUG_ON to do
ordering checking.
This patch add alloc_low_pages to make it possible to allocate serveral
pages at first, and hand out pages one by one from low to high.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-28-git-send-email-yinghai@kernel.org
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Page table area are pre-mapped now after
x86, mm: setup page table in top-down
x86, mm: Remove early_memremap workaround for page table accessing on 64bit
mapping_pagetable_reserve is not used anymore, so remove it.
Also remove operation in mask_rw_pte(), as modified allow_low_page
always return pages that are already mapped, moreover
xen_alloc_pte_init, xen_alloc_pmd_init, etc, will mark the page RO
before hooking it into the pagetable automatically.
-v2: add changelog about mask_rw_pte() from Stefano.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-27-git-send-email-yinghai@kernel.org
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
They are almost same except 64 bit need to handle after_bootmem case.
Add mm_internal.h to make that alloc_low_page() only to be accessible
from arch/x86/mm/init*.c
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-25-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Now all page table buf are pre-mapped, and could use virtual address directly.
So don't need to remember physical address anymore.
Remove that phys pointer in alloc_low_page(), and that will allow us to merge
alloc_low_page between 64bit and 32bit.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-24-git-send-email-yinghai@kernel.org
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We try to put page table high to make room for kdump, and at that time
those ranges are not mapped yet, and have to use ioremap to access it.
Now after patch that pre-map page table top down.
x86, mm: setup page table in top-down
We do not need that workaround anymore.
Just use __va to return directly mapping address.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-23-git-send-email-yinghai@kernel.org
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Get pgt_buf early from BRK, and use it to map PMD_SIZE from top at first.
Then use mapped pages to map more ranges below, and keep looping until
all pages get mapped.
alloc_low_page will use page from BRK at first, after that buffer is used
up, will use memblock to find and reserve pages for page table usage.
Introduce min_pfn_mapped to make sure find new pages from mapped ranges,
that will be updated when lower pages get mapped.
Also add step_size to make sure that don't try to map too big range with
limited mapped pages initially, and increase the step_size when we have
more mapped pages on hand.
We don't need to call pagetable_reserve anymore, reserve work is done
in alloc_low_page() directly.
At last we can get rid of calculation and find early pgt related code.
-v2: update to after fix_xen change,
also use MACRO for initial pgt_buf size and add comments with it.
-v3: skip big reserved range in memblock.reserved near end.
-v4: don't need fix_xen change now.
-v5: add changelog about moving about reserving pagetable to alloc_low_page.
Suggested-by: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-22-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
After we add code use buffer in BRK to pre-map buf for page table in
following patch:
x86, mm: setup page table in top-down
it should be safe to remove early_memmap for page table accessing.
Instead we get panic with that.
It turns out that we clear the initial page table wrongly for next range
that is separated by holes.
And it only happens when we are trying to map ram range one by one.
We need to check if the range is ram before clearing page table.
We change the loop structure to remove the extra little loop and use
one loop only, and in that loop will caculate next at first, and check if
[addr,next) is covered by E820_RAM.
-v2: E820_RESERVED_KERN is treated as E820_RAM. EFI one change some E820_RAM
to that, so next kernel by kexec will know that range is used already.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-20-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We could map small range in the middle of big range at first, so should use
big page size at first to avoid using small page size to break down page table.
Only can set big page bit when that range has ram area around it.
-v2: fix 32bit boundary checking. We can not count ram above max_low_pfn
for 32 bit.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-19-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We are going to use buffer in BRK to map small range just under memory top,
and use those new mapped ram to map ram range under it.
The ram range that will be mapped at first could be only page aligned,
but ranges around it are ram too, we could use bigger page to map it to
avoid small page size.
We will adjust page_size_mask in following patch:
x86, mm: Use big page size for small memory range
to use big page size for small ram range.
Before that patch, this patch will make sure start address to be
aligned down according to bigger page size, otherwise entry in page
page will not have correct value.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-18-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Currently direct mappings are created for [ 0 to max_low_pfn<<PAGE_SHIFT )
and [ 4GB to max_pfn<<PAGE_SHIFT ), which may include regions that are not
backed by actual DRAM. This is fine for holes under 4GB which are covered
by fixed and variable range MTRRs to be UC. However, we run into trouble
on higher memory addresses which cannot be covered by MTRRs.
Our system with 1TB of RAM has an e820 that looks like this:
BIOS-e820: [mem 0x0000000000000000-0x00000000000983ff] usable
BIOS-e820: [mem 0x0000000000098400-0x000000000009ffff] reserved
BIOS-e820: [mem 0x00000000000d0000-0x00000000000fffff] reserved
BIOS-e820: [mem 0x0000000000100000-0x00000000c7ebffff] usable
BIOS-e820: [mem 0x00000000c7ec0000-0x00000000c7ed7fff] ACPI data
BIOS-e820: [mem 0x00000000c7ed8000-0x00000000c7ed9fff] ACPI NVS
BIOS-e820: [mem 0x00000000c7eda000-0x00000000c7ffffff] reserved
BIOS-e820: [mem 0x00000000fec00000-0x00000000fec0ffff] reserved
BIOS-e820: [mem 0x00000000fee00000-0x00000000fee00fff] reserved
BIOS-e820: [mem 0x00000000fff00000-0x00000000ffffffff] reserved
BIOS-e820: [mem 0x0000000100000000-0x000000e037ffffff] usable
BIOS-e820: [mem 0x000000e038000000-0x000000fcffffffff] reserved
BIOS-e820: [mem 0x0000010000000000-0x0000011ffeffffff] usable
and so direct mappings are created for huge memory hole between
0x000000e038000000 to 0x0000010000000000. Even though the kernel never
generates memory accesses in that region, since the page tables mark
them incorrectly as being WB, our (AMD) processor ends up causing a MCE
while doing some memory bookkeeping/optimizations around that area.
This patch iterates through e820 and only direct maps ranges that are
marked as E820_RAM, and keeps track of those pfn ranges. Depending on
the alignment of E820 ranges, this may possibly result in using smaller
size (i.e. 4K instead of 2M or 1G) page tables.
-v2: move changes from setup.c to mm/init.c, also use for_each_mem_pfn_range
instead. - Yinghai Lu
-v3: add calculate_all_table_space_size() to get correct needed page table
size. - Yinghai Lu
-v4: fix add_pfn_range_mapped() to get correct max_low_pfn_mapped when
mem map does have hole under 4g that is found by Konard on xen
domU with 8g ram. - Yinghai
Signed-off-by: Jacob Shin <jacob.shin@amd.com>
Link: http://lkml.kernel.org/r/1353123563-3103-16-git-send-email-yinghai@kernel.org
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Reviewed-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We are going to map ram only, so under max_low_pfn_mapped,
between 4g and max_pfn_mapped does not mean mapped at all.
Use pfn_range_is_mapped() to find out if range is mapped for initrd.
That could happen bootloader put initrd in range but user could
use memmap to carve some of range out.
Also during copying need to use early_memmap to map original initrd
for accessing.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-15-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We are going to map ram only, so under max_low_pfn_mapped,
between 4g and max_pfn_mapped does not mean mapped at all.
Use pfn_range_is_mapped() directly.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-14-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We are going to map ram only, so under max_low_pfn_mapped,
between 4g and max_pfn_mapped does not mean mapped at all.
Use pfn_range_is_mapped() directly.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-13-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Update code that previously assumed pfns [ 0 - max_low_pfn_mapped ) and
[ 4GB - max_pfn_mapped ) were always direct mapped, to now look up
pfn_mapped ranges instead.
-v2: change applying sequence to keep git bisecting working.
so add dummy pfn_range_is_mapped(). - Yinghai Lu
Signed-off-by: Jacob Shin <jacob.shin@amd.com>
Link: http://lkml.kernel.org/r/1353123563-3103-12-git-send-email-yinghai@kernel.org
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
There could be cases where user supplied memmap=exactmap memory
mappings do not mark the region where the kernel .text .data and
.bss reside as E820_RAM, as reported here:
https://lkml.org/lkml/2012/8/14/86
Handle it by complaining, and adding the range back into the e820.
Signed-off-by: Jacob Shin <jacob.shin@amd.com>
Link: http://lkml.kernel.org/r/1353123563-3103-11-git-send-email-yinghai@kernel.org
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Reviewed-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
memblock_x86_fill() could double memory array.
If we set memblock.current_limit to 512M, so memory array could be around 512M.
So kdump will not get big range (like 512M) under 1024M.
Try to put it down under 1M, it would use about 4k or so, and that is limited.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-10-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
It should take physical address range that will need to be mapped.
find_early_table_space should take range that pgt buff should be in.
Separating page table size calculating and finding early page table to
reduce confusing.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-9-git-send-email-yinghai@kernel.org
Reviewed-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
We should not do that in every calling of init_memory_mapping.
At the same time need to move down early_memtest, and could remove after_bootmem
checking.
-v2: fix one early_memtest with 32bit by passing max_pfn_mapped instead.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-8-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
After
| commit 8548c84da2
| Author: Takashi Iwai <tiwai@suse.de>
| Date: Sun Oct 23 23:19:12 2011 +0200
|
| x86: Fix S4 regression
|
| Commit 4b239f458 ("x86-64, mm: Put early page table high") causes a S4
| regression since 2.6.39, namely the machine reboots occasionally at S4
| resume. It doesn't happen always, overall rate is about 1/20. But,
| like other bugs, once when this happens, it continues to happen.
|
| This patch fixes the problem by essentially reverting the memory
| assignment in the older way.
Have some page table around 512M again, that will prevent kdump to find 512M
under 768M.
We need revert that reverting, so we could put page table high again for 64bit.
Takashi agreed that S4 regression could be something else.
https://lkml.org/lkml/2012/6/15/182
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-6-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Now init_memory_mapping is called two times, later will be called for every
ram ranges.
Could put all related init_mem calling together and out of setup.c.
Actually, it reverts commit 1bbbbe7
x86: Exclude E820_RESERVED regions and memory holes above 4 GB from direct mapping.
will address that later with complete solution include handling hole under 4g.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-5-git-send-email-yinghai@kernel.org
Reviewed-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
So make init_memory_mapping smaller and readable.
-v2: use 0 instead of nr_range as input parameter found by Yasuaki Ishimatsu.
Suggested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-3-git-send-email-yinghai@kernel.org
Reviewed-by: Pekka Enberg <penberg@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Now we pass around use_gbpages and use_pse for calculating page table size,
Later we will need to call init_memory_mapping for every ram range one by one,
that mean those calculation will be done several times.
Those information are the same for all ram range and could be stored in
page_size_mask and could be probed it one time only.
Move that probing code out of init_memory_mapping into separated function
probe_page_size_mask(), and call it before all init_memory_mapping.
Suggested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1353123563-3103-2-git-send-email-yinghai@kernel.org
Reviewed-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Just use BUG_ON() instead of constructions such as:
if (...)
BUG()
A simplified version of the semantic patch that makes this transformation
is as follows: (http://coccinelle.lip6.fr/)
// <smpl>
@@
expression e;
@@
- if (e) BUG();
+ BUG_ON(e);
// </smpl>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This pushes the dependencies on the Integrator/AP system
controller (SC) down into the PCI V3 driver and the
AP-specific board file.
First, the platform data for the PL010 UART is moved into
the integrator_ap.c board file, and the Integrator/CP is
assigned with NULL pdata. This way the callback functions
can reference the dynamically remapped AP syscon address
in both the ATAG and DT boot path, and this remapping
is localized to the board file.
Second the PCIv3 driver is making its own dynamic remapping
of the SC for the few registers it is using. When we
convert the PCIv3 driver over to using device tree having a
dynamically assigned base address will be useful, but we
will have to use the definition from <mach/platform.h> for
now, the only improvement is that it's done dynamically.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
If we are in nohz and there's still irq_work to be done when the idle
task is about to go offline, give a nasty warning. Everything should
have been flushed from the CPU_DYING notifier already. Further attempts
to enqueue an irq_work are buggy because irqs are disabled by
__cpu_disable(). The best we can do is to report the issue to the user.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
In order not to offline a CPU with pending irq works, flush the
queue from CPU_DYING. The notifier is called by stop_machine on
the CPU that is going down. The code will not be called from irq context
(so things like get_irq_regs() wont work) but I'm not sure what the
requirements are for irq_work in that regard (Peter?). But irqs are
disabled and the CPU is about to go offline. Might as well flush the work.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Don't stop the tick if we have pending irq works on the
queue, otherwise if the arch can't raise self-IPIs, we may not
find an opportunity to execute the pending works for a while.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
We need some quick way to check if the CPU has stopped
its tick. This will be useful to implement the printk tick
using the irq work subsystem.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
irq work can run on any arch even without IPI
support because of the hook on update_process_times().
So lets remove HAVE_IRQ_WORK because it doesn't reflect
any backend requirement.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Error handling in xfs_buf_ioapply_map() does not handle IO reference
counts correctly. We increment the b_io_remaining count before
building the bio, but then fail to decrement it in the failure case.
This leads to the buffer never running IO completion and releasing
the reference that the IO holds, so at unmount we can leak the
buffer. This leak is captured by this assert failure during unmount:
XFS: Assertion failed: atomic_read(&pag->pag_ref) == 0, file: fs/xfs/xfs_mount.c, line: 273
This is not a new bug - the b_io_remaining accounting has had this
problem for a long, long time - it's just very hard to get a
zero length bio being built by this code...
Further, the buffer IO error can be overwritten on a multi-segment
buffer by subsequent bio completions for partial sections of the
buffer. Hence we should only set the buffer error status if the
buffer is not already carrying an error status. This ensures that a
partial IO error on a multi-segment buffer will not be lost. This
part of the problem is a regression, however.
cc: <stable@vger.kernel.org>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
When we shut down the filesystem, it might first be detected in
writeback when we are allocating a inode size transaction. This
happens after we have moved all the pages into the writeback state
and unlocked them. Unfortunately, if we fail to set up the
transaction we then abort writeback and try to invalidate the
current page. This then triggers are BUG() in block_invalidatepage()
because we are trying to invalidate an unlocked page.
Fixing this is a bit of a chicken and egg problem - we can't
allocate the transaction until we've clustered all the pages into
the IO and we know the size of it (i.e. whether the last block of
the IO is beyond the current EOF or not). However, we don't want to
hold pages locked for long periods of time, especially while we lock
other pages to cluster them into the write.
To fix this, we need to make a clear delineation in writeback where
errors can only be handled by IO completion processing. That is,
once we have marked a page for writeback and unlocked it, we have to
report errors via IO completion because we've already started the
IO. We may not have submitted any IO, but we've changed the page
state to indicate that it is under IO so we must now use the IO
completion path to report errors.
To do this, add an error field to xfs_submit_ioend() to pass it the
error that occurred during the building on the ioend chain. When
this is non-zero, mark each ioend with the error and call
xfs_finish_ioend() directly rather than building bios. This will
immediately push the ioends through completion processing with the
error that has occurred.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
In certain circumstances, a double split of an attribute tree is
needed to insert or replace an attribute. In rare situations, this
can go wrong, leaving the attribute tree corrupted. In this case,
the attr being replaced is the last attr in a leaf node, and the
replacement is larger so doesn't fit in the same leaf node.
When we have the initial condition of a node format attribute
btree with two leaves at index 1 and 2. Call them L1 and L2. The
leaf L1 is completely full, there is not a single byte of free space
in it. L2 is mostly empty. The attribute being replaced - call it X
- is the last attribute in L1.
The way an attribute replace is executed is that the replacement
attribute - call it Y - is first inserted into the tree, but has an
INCOMPLETE flag set on it so that list traversals ignore it. Once
this transaction is committed, a second transaction it run to
atomically mark Y as COMPLETE and X as INCOMPLETE, so that a
traversal will now find Y and skip X. Once that transaction is
committed, attribute X is then removed.
So, the initial condition is:
+--------+ +--------+
| L1 | | L2 |
| fwd: 2 |---->| fwd: 0 |
| bwd: 0 |<----| bwd: 1 |
| fsp: 0 | | fsp: N |
|--------| |--------|
| attr A | | attr 1 |
|--------| |--------|
| attr B | | attr 2 |
|--------| |--------|
.......... ..........
|--------| |--------|
| attr X | | attr n |
+--------+ +--------+
So now we go to replace X, and see that L1:fsp = 0 - it is full so
we can't insert Y in the same leaf. So we record the the location of
attribute X so we can track it for later use, then we split L1 into
L1 and L3 and reblance across the two leafs. We end with:
+--------+ +--------+ +--------+
| L1 | | L3 | | L2 |
| fwd: 3 |---->| fwd: 2 |---->| fwd: 0 |
| bwd: 0 |<----| bwd: 1 |<----| bwd: 3 |
| fsp: M | | fsp: J | | fsp: N |
|--------| |--------| |--------|
| attr A | | attr X | | attr 1 |
|--------| +--------+ |--------|
| attr B | | attr 2 |
|--------| |--------|
.......... ..........
|--------| |--------|
| attr W | | attr n |
+--------+ +--------+
And we track that the original attribute is now at L3:0.
We then try to insert Y into L1 again, and find that there isn't
enough room because the new attribute is larger than the old one.
Hence we have to split again to make room for Y. We end up with
this:
+--------+ +--------+ +--------+ +--------+
| L1 | | L4 | | L3 | | L2 |
| fwd: 4 |---->| fwd: 3 |---->| fwd: 2 |---->| fwd: 0 |
| bwd: 0 |<----| bwd: 1 |<----| bwd: 4 |<----| bwd: 3 |
| fsp: M | | fsp: J | | fsp: J | | fsp: N |
|--------| |--------| |--------| |--------|
| attr A | | attr Y | | attr X | | attr 1 |
|--------| + INCOMP + +--------+ |--------|
| attr B | +--------+ | attr 2 |
|--------| |--------|
.......... ..........
|--------| |--------|
| attr W | | attr n |
+--------+ +--------+
And now we have the new (incomplete) attribute @ L4:0, and the
original attribute at L3:0. At this point, the first transaction is
committed, and we move to the flipping of the flags.
This is where we are supposed to end up with this:
+--------+ +--------+ +--------+ +--------+
| L1 | | L4 | | L3 | | L2 |
| fwd: 4 |---->| fwd: 3 |---->| fwd: 2 |---->| fwd: 0 |
| bwd: 0 |<----| bwd: 1 |<----| bwd: 4 |<----| bwd: 3 |
| fsp: M | | fsp: J | | fsp: J | | fsp: N |
|--------| |--------| |--------| |--------|
| attr A | | attr Y | | attr X | | attr 1 |
|--------| +--------+ + INCOMP + |--------|
| attr B | +--------+ | attr 2 |
|--------| |--------|
.......... ..........
|--------| |--------|
| attr W | | attr n |
+--------+ +--------+
But that doesn't happen properly - the attribute tracking indexes
are not pointing to the right locations. What we end up with is both
the old attribute to be removed pointing at L4:0 and the new
attribute at L4:1. On a debug kernel, this assert fails like so:
XFS: Assertion failed: args->index2 < be16_to_cpu(leaf2->hdr.count), file: fs/xfs/xfs_attr_leaf.c, line: 2725
because the new attribute location does not exist. On a production
kernel, this goes unnoticed and the code proceeds ahead merrily and
removes L4 because it thinks that is the block that is no longer
needed. This leaves the hash index node pointing to entries
L1, L4 and L2, but only blocks L1, L3 and L2 to exist. Further, the
leaf level sibling list is L1 <-> L4 <-> L2, but L4 is now free
space, and so everything is busted. This corruption is caused by the
removal of the old attribute triggering a join - it joins everything
correctly but then frees the wrong block.
xfs_repair will report something like:
bad sibling back pointer for block 4 in attribute fork for inode 131
problem with attribute contents in inode 131
would clear attr fork
bad nblocks 8 for inode 131, would reset to 3
bad anextents 4 for inode 131, would reset to 0
The problem lies in the assignment of the old/new blocks for
tracking purposes when the double leaf split occurs. The first split
tries to place the new attribute inside the current leaf (i.e.
"inleaf == true") and moves the old attribute (X) to the new block.
This sets up the old block/index to L1:X, and newly allocated
block to L3:0. It then moves attr X to the new block and tries to
insert attr Y at the old index. That fails, so it splits again.
With the second split, the rebalance ends up placing the new attr in
the second new block - L4:0 - and this is where the code goes wrong.
What is does is it sets both the new and old block index to the
second new block. Hence it inserts attr Y at the right place (L4:0)
but overwrites the current location of the attr to replace that is
held in the new block index (currently L3:0). It over writes it with
L4:1 - the index we later assert fail on.
Hopefully this table will show this in a foramt that is a bit easier
to understand:
Split old attr index new attr index
vanilla patched vanilla patched
before 1st L1:26 L1:26 N/A N/A
after 1st L3:0 L3:0 L1:26 L1:26
after 2nd L4:0 L3:0 L4:1 L4:0
^^^^ ^^^^
wrong wrong
The fix is surprisingly simple, for all this analysis - just stop
the rebalance on the out-of leaf case from overwriting the new attr
index - it's already correct for the double split case.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
We can't assume this device exists, fall back to the bridge itself.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Tested-by: Matthew Thode <prometheanfire@gentoo.org>
Cc: stable@vger.kernel.org
Signed-off-by: Joerg Roedel <joro@8bytes.org>
Add support for debug_dma_mapping_error() call to avoid warning from
debug_dma_unmap() interface when it checks for mapping error checked
status. Without this patch, device driver failed to check map error
warning is generated.
Signed-off-by: Shuah Khan <shuah.khan@hp.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Joerg Roedel <joro@8bytes.org>
dma-debug depends on get_dma_ops() interface. Several architectures
do not define dma_ops and get_dma_ops(). When dma debug interfaces are
used on an architecture (e.g: c6x) that doesn't define get_dmap_ops(),
compilation fails. Changing dma-debug to call dma_mapping_error() instead
of defining its own that calls get_dma_ops(), such that the internal use of
dma_mapping_error() doesn't interfere with the debug_dma_mapping_error()
interface's mapping error checks. Moving dma_mapping_error() checks in
check_unmap() under the dma debug entry not found is sufficient to fix the
problem.
Reference: https://lkml.org/lkml/2012/10/26/367
Signed-off-by: Shuah Khan <shuah.khan@hp.com>
Reported-by: Mark Salter <msalter@redhat.com>
Signed-off-by: Joerg Roedel <joro@8bytes.org>
Neither of these should ever be changed once set. Make them const and
fix up the users that try to modify it in-place. In one case
kmalloc+memcpy is replaced with kstrdup() to avoid modifying the string.
Build tested with defconfigs on ARM, PowerPC, Sparc, MIPS, x86 among
others.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Julian Calaby <julian.calaby@gmail.com>