The local variable speed does not need to be initialized and can cause some
static analysis tools to complain the initial assigned value is never used.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add few checks to validate msg from iavf driver.
Test if we have got enough q_vectors allocated in VSI connected with VF.
Add masks for itr_indx and msix_indx to avoid writing to reserved fieldi
of QINT. Clear q_vector->num_ring_rx/tx, without it we can increment this
value every time we send irq map msg from VF. So after second call this
value will be incorrect.
Decrement num_vectors from msg, because last vector in iavf msg is misc
vector (we don't set map for it).
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
In case of non-trusted VFs, it is possible to program VLAN filter far
less than what is requested by the VF originally, thereby makes number of
VLAN elements being tracked by VF different from actual VLAN tags. This
patch makes sure that we are not attempting to remove VLAN filter that
does not exist.
Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When Tx insertion is set, we are not accounting for the state of Rx
stripping. This causes Rx stripping to be enabled any time Tx
insertion is changed, even when it's supposed to be disabled.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Disable VF if any malicious device driver (MDD) event is detected by
hardware. Track vf->num_mdd_events for information about VF MDD events.
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Many times in our functions we have a local variable pf, which is
equivalent to vsi->back. Just use pf consistently instead of vsi->back
where available.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
One regression fix.
Changes we merged to STRICT_KERNEL_RWX on 32-bit were causing crashes under
load on some machines depending on memory layout.
Thanks to:
Christophe Leroy.
-----BEGIN PGP SIGNATURE-----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=/m4Q
-----END PGP SIGNATURE-----
Merge tag 'powerpc-5.1-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fix from Michael Ellerman:
"One regression fix.
Changes we merged to STRICT_KERNEL_RWX on 32-bit were causing crashes
under load on some machines depending on memory layout.
Thanks to Christophe Leroy"
* tag 'powerpc-5.1-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/32s: Fix BATs setting with CONFIG_STRICT_KERNEL_RWX
This patch adds the implementation of the HL_DEBUG_OP_SET_MODE opcode in
the DEBUG IOCTL.
It forces the user who wants to debug the device to set the device into
debug mode before he can configure the debug engines. The patch also makes
sure to disable debug mode upon user releasing FD, in case the user forgot
to disable debug mode.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This patch fixes comments on various structure members and some spelling
errors in log messages.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This patch removes redundant CPU availability checks in:
goya_test_queues() - will be done in goya_test_cpu_queue().
goya_ring_doorbell() - was done earlier in goya_send_cpu_message().
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This patch improves the error message that is shown when a new user tries
to open a new FD while there is already an existing user that is working
on the device.
It also improves the error message in case of missing firmware file.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Now freeing hw queue resource is moved to hctx's release handler,
we don't need to worry about the race between blk_cleanup_queue and
run queue any more.
So don't drain in-progress dispatch in blk_cleanup_queue().
This is basically revert of c2856ae2f3 ("blk-mq: quiesce queue before
freeing queue").
Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: linux-scsi@vger.kernel.org,
Cc: Martin K . Petersen <martin.petersen@oracle.com>,
Cc: Christoph Hellwig <hch@lst.de>,
Cc: James E . J . Bottomley <jejb@linux.vnet.ibm.com>,
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Tested-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
hctx is always released after requeue is freed.
With holding queue's kobject refcount, it is safe for driver to run queue,
so one run queue might be scheduled after blk_sync_queue() is done.
So moving the cancel of hctx->run_work into blk_mq_hw_sysfs_release()
for avoiding run released queue.
Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: linux-scsi@vger.kernel.org,
Cc: Martin K . Petersen <martin.petersen@oracle.com>,
Cc: Christoph Hellwig <hch@lst.de>,
Cc: James E . J . Bottomley <jejb@linux.vnet.ibm.com>,
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Tested-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In normal queue cleanup path, hctx is released after request queue
is freed, see blk_mq_release().
However, in __blk_mq_update_nr_hw_queues(), hctx may be freed because
of hw queues shrinking. This way is easy to cause use-after-free,
because: one implicit rule is that it is safe to call almost all block
layer APIs if the request queue is alive; and one hctx may be retrieved
by one API, then the hctx can be freed by blk_mq_update_nr_hw_queues();
finally use-after-free is triggered.
Fixes this issue by always freeing hctx after releasing request queue.
If some hctxs are removed in blk_mq_update_nr_hw_queues(), introduce
a per-queue list to hold them, then try to resuse these hctxs if numa
node is matched.
Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: linux-scsi@vger.kernel.org,
Cc: Martin K . Petersen <martin.petersen@oracle.com>,
Cc: Christoph Hellwig <hch@lst.de>,
Cc: James E . J . Bottomley <jejb@linux.vnet.ibm.com>,
Reviewed-by: Hannes Reinecke <hare@suse.com>
Tested-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Split blk_mq_alloc_and_init_hctx into two parts, and one is
blk_mq_alloc_hctx() for allocating all hctx resources, another
is blk_mq_init_hctx() for initializing hctx, which serves as
counter-part of blk_mq_exit_hctx().
Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: linux-scsi@vger.kernel.org
Cc: Martin K . Petersen <martin.petersen@oracle.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James E . J . Bottomley <jejb@linux.vnet.ibm.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Once blk_cleanup_queue() returns, tags shouldn't be used any more,
because blk_mq_free_tag_set() may be called. Commit 45a9c9d909
("blk-mq: Fix a use-after-free") fixes this issue exactly.
However, that commit introduces another issue. Before 45a9c9d909,
we are allowed to run queue during cleaning up queue if the queue's
kobj refcount is held. After that commit, queue can't be run during
queue cleaning up, otherwise oops can be triggered easily because
some fields of hctx are freed by blk_mq_free_queue() in blk_cleanup_queue().
We have invented ways for addressing this kind of issue before, such as:
8dc765d438 ("SCSI: fix queue cleanup race before queue initialization is done")
c2856ae2f3 ("blk-mq: quiesce queue before freeing queue")
But still can't cover all cases, recently James reports another such
kind of issue:
https://marc.info/?l=linux-scsi&m=155389088124782&w=2
This issue can be quite hard to address by previous way, given
scsi_run_queue() may run requeues for other LUNs.
Fixes the above issue by freeing hctx's resources in its release handler, and this
way is safe becasue tags isn't needed for freeing such hctx resource.
This approach follows typical design pattern wrt. kobject's release handler.
Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: linux-scsi@vger.kernel.org,
Cc: Martin K . Petersen <martin.petersen@oracle.com>,
Cc: Christoph Hellwig <hch@lst.de>,
Cc: James E . J . Bottomley <jejb@linux.vnet.ibm.com>,
Reported-by: James Smart <james.smart@broadcom.com>
Fixes: 45a9c9d909 ("blk-mq: Fix a use-after-free")
Cc: stable@vger.kernel.org
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
With holding queue's kobject refcount, it is safe for driver
to schedule requeue. However, blk_mq_kick_requeue_list() may
be called after blk_sync_queue() is done because of concurrent
requeue activities, then requeue work may not be completed when
freeing queue, and kernel oops is triggered.
So moving the cancel of requeue_work into blk_mq_release() for
avoiding race between requeue and freeing queue.
Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: linux-scsi@vger.kernel.org,
Cc: Martin K . Petersen <martin.petersen@oracle.com>,
Cc: Christoph Hellwig <hch@lst.de>,
Cc: James E . J . Bottomley <jejb@linux.vnet.ibm.com>,
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Just like aio/io_uring, we need to grab 2 refcount for queuing one
request, one is for submission, another is for completion.
If the request isn't queued from plug code path, the refcount grabbed
in generic_make_request() serves for submission. In theroy, this
refcount should have been released after the sumission(async run queue)
is done. blk_freeze_queue() works with blk_sync_queue() together
for avoiding race between cleanup queue and IO submission, given async
run queue activities are canceled because hctx->run_work is scheduled with
the refcount held, so it is fine to not hold the refcount when
running the run queue work function for dispatch IO.
However, if request is staggered into plug list, and finally queued
from plug code path, the refcount in submission side is actually missed.
And we may start to run queue after queue is removed because the queue's
kobject refcount isn't guaranteed to be grabbed in flushing plug list
context, then kernel oops is triggered, see the following race:
blk_mq_flush_plug_list():
blk_mq_sched_insert_requests()
insert requests to sw queue or scheduler queue
blk_mq_run_hw_queue
Because of concurrent run queue, all requests inserted above may be
completed before calling the above blk_mq_run_hw_queue. Then queue can
be freed during the above blk_mq_run_hw_queue().
Fixes the issue by grab .q_usage_counter before calling
blk_mq_sched_insert_requests() in blk_mq_flush_plug_list(). This way is
safe because the queue is absolutely alive before inserting request.
Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: linux-scsi@vger.kernel.org,
Cc: Martin K . Petersen <martin.petersen@oracle.com>,
Cc: Christoph Hellwig <hch@lst.de>,
Cc: James E . J . Bottomley <jejb@linux.vnet.ibm.com>,
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Tested-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Status of ADMA channel registers is not saved and restored during system
suspend. During active playback if system enters suspend, this results in
wrong state of channel registers during system resume and playback fails
to resume properly. Fix this by saving following channel registers in
runtime suspend and restore during runtime resume.
* ADMA_CH_LOWER_SRC_ADDR
* ADMA_CH_LOWER_TRG_ADDR
* ADMA_CH_FIFO_CTRL
* ADMA_CH_CONFIG
* ADMA_CH_CTRL
* ADMA_CH_CMD
* ADMA_CH_TC
Runtime PM calls will be inovked during system resume path if a playback
or capture needs to be resumed. Hence above changes work fine for system
suspend case.
Fixes: f46b195799 ("dmaengine: tegra-adma: Add support for Tegra210 ADMA")
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
During an audio playback session it is observed that, audio goes off after
few seconds of continuous pause and play. No audio is heard even when the
playback is resumed.
The reason for above is, currently ADMA driver does not handle DMA_PAUSE/
DMA_RESUME and relevant callbacks for dma_device are not implemented. This
patch implements device_pause and device_resume callbacks for the device.
During pause TRANSFER_PAUSE bit of dma channel control register is set and
the same is cleared during resume.
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add Tegra186 specific macro defines and chip_data structure for chip
specific information. New compatibility is added to select relevant
chip details. There is no major change for Tegra194 and hence it can
use the same chip data.
The bits in the BURST_SIZE field of the ADMA CH_CONFIG register are
encoded differently on Tegra186 and Tegra194 compared with Tegra210.
On Tegra210 the bits are encoded as follows ...
1 = WORD_1
2 = WORDS_2
3 = WORDS_4
4 = WORDS_8
5 = WORDS_16
Where as on Tegra186 and Tegra194 the bits are encoded as ...
0 = WORD_1
1 = WORDS_2
2 = WORDS_3
3 = WORDS_4
4 = WORDS_5
...
15 = WORDS_16
Add helper functions for generating the correct burst size.
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Tegra186 Audio DMA controller has new updates from Tegra210 version.
Thus add new compatibility binding string and the same can be used
by Tegra194 as well.
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
This is a preparatory patch to add support for Tegra186 and Tegra194 chips.
Following changes are necessary to make driver code generic.
* chip_data structure is enhanced to have chip specific details and
following are the additions to the structure
* Offset addresses for ADMA global and channel registers
* Offset values for Tx and Rx channel selection
* Maximum supported Tx and Rx channels
* Tx and Rx channel request mask
* ADMA channel register space size
* Make use of above chip_data to generalise the driver code
Support for Tegra186 and Tegra194 will be added in subsequent patches of
the series.
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
We switched this code from spin_lock_bh() to vanilla spin_lock() but
there was one stray spin_unlock_bh() that was overlooked. This
patch converts it to spin_unlock() as well.
Fixes: d8570d018f ("dmaengine: at_xdmac: move spin_lock_bh to spin_lock in tasklet")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Fix indentation and remove unneeded space after 'return' keyword. This
fixes checkpatch warning:
WARNING: Statements should start on a tabstop
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
In double buffer mode, during residue calculation, the DMA can
automatically switch to the next transfer. Indeed the CT bit that
gives position in the double buffer can has been updated by the
hardware, during calculation.
In this case the SxNDTR register value can not be trusted.
If a transition is detected we consider that the DMA has switched to
the beginning of next sg.
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
Signed-off-by: Pierre-Yves MORDRET <pierre-yves.mordret@st.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Asking the GPU to busywait on a memory address, perhaps not unexpectedly
in hindsight for a shared system, leads to bus contention that affects
CPU programs trying to concurrently access memory. This can manifest as
a drop in transcode throughput on highly over-saturated workloads.
The only clue offered by perf, is that the bus-cycles (perf stat -e
bus-cycles) jumped by 50% when enabling semaphores. This corresponds
with extra CPU active cycles being attributed to intel_idle's mwait.
This patch introduces a heuristic to try and detect when more than one
client is submitting to the GPU pushing it into an oversaturated state.
As we already keep track of when the semaphores are signaled, we can
inspect their state on submitting the busywait batch and if we planned
to use a semaphore but were too late, conclude that the GPU is
overloaded and not try to use semaphores in future requests. In
practice, this means we optimistically try to use semaphores for the
first frame of a transcode job split over multiple engines, and fail if
there are multiple clients active and continue not to use semaphores for
the subsequent frames in the sequence. Periodically, we try to
optimistically switch semaphores back on whenever the client waits to
catch up with the transcode results.
With 1 client, on Broxton J3455, with the relative fps normalized by %cpu:
x no semaphores
+ drm-tip
* patched
+------------------------------------------------------------------------+
| * |
| *+ |
| **+ |
| **+ x |
| x * +**+ x |
| x x * * +***x xx |
| x x * * *+***x *x |
| x x* + * * *****x *x x |
| + x xx+x* + *** * ********* x * |
| + x xx+x* * *** +** ********* xx * |
| * + ++++* + x*x****+*+* ***+*************+x* * |
|*+ +** *+ + +* + *++****** *xxx**********x***+*****************+*++ *|
| |__________A_____M_____| |
| |_______________A____M_________| |
| |____________A___M________| |
+------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 120 2.60475 3.50941 3.31123 3.2143953 0.21117399
+ 120 2.3826 3.57077 3.25101 3.1414161 0.28146407
Difference at 95.0% confidence
-0.0729792 +/- 0.0629585
-2.27039% +/- 1.95864%
(Student's t, pooled s = 0.248814)
* 120 2.35536 3.66713 3.2849 3.2059917 0.24618565
No difference proven at 95.0% confidence
With 10 clients over-saturating the pipeline:
x no semaphores
+ drm-tip
* patched
+------------------------------------------------------------------------+
| ++ ** |
| ++ ** |
| ++ ** |
| ++ ** |
| ++ xx *** |
| ++ xx *** |
| ++ xxx*** |
| ++ xxx*** |
| +++ xxx*** |
| +++ xx**** |
| +++ xx**** |
| +++ xx**** |
| +++ xx**** |
| ++++ xx**** |
| +++++ xx**** |
| +++++ x x****** |
| ++++++ xxx******* |
| ++++++ xxx******* |
| ++++++ xxx******* |
| ++++++ xx******** |
| ++++++ xxxx******** |
| ++++++ xxxx******** |
| ++++++++ xxxxx********* |
|+ + + + ++++++++ xxx*xx**********x* *|
| |__A__| |
| |__AM__| |
| |__A_| |
+------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 120 2.47855 2.8972 2.72376 2.7193402 0.074604933
+ 120 1.17367 1.77459 1.71977 1.6966782 0.085850697
Difference at 95.0% confidence
-1.02266 +/- 0.0203502
-37.607% +/- 0.748352%
(Student's t, pooled s = 0.0804246)
* 120 2.57868 3.00821 2.80142 2.7923878 0.058646477
Difference at 95.0% confidence
0.0730476 +/- 0.0169791
2.68622% +/- 0.624383%
(Student's t, pooled s = 0.0671018)
Indicating that we've recovered the regression from enabling semaphores
on this saturated setup, with a hint towards an overall improvement.
Very similar, but of smaller magnitude, results are observed on both
Skylake(gt2) and Kabylake(gt4). This may be due to the reduced impact of
bus-cycles, where we see a 50% hit on Broxton, it is only 10% on the big
core, in this particular test.
One observation to make here is that for a greedy client trying to
maximise its own throughput, using semaphores is the right choice. It is
only the holistic system-wide view that semaphores of one client
impacts another and reduces the overall throughput where we would choose
to disable semaphores.
The most noticeable negactive impact this has is on the no-op
microbenchmarks, which are also very notable for having no cpu bus load.
In particular, this increases the runtime and energy consumption of
gem_exec_whisper.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
Cc: Dmitry Ermilov <dmitry.ermilov@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190504070707.30902-1-chris@chris-wilson.co.uk
e is the counter used to save the location of a dump when an
skb is filled. Once the walk of the table is complete, mr_table_dump
needs to return without resetting that index to 0. Dump of a specific
table is looping because of the reset because there is no way to
indicate the walk of the table is done.
Move the reset to the caller so the dump of each table starts at 0,
but the loop counter is maintained if a dump fills an skb.
Fixes: e1cedae1ba ("ipmr: Refactor mr_rtm_dumproute")
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For all other error cases in queue_userspace_packet() the error is
returned, so it makes sense to do the same for these two error cases.
Reported-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
rtl_write_exgmac_batch is used in only one place, so we can remove it.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Michal Kubecek says:
====================
netlink: strict attribute checking follow-up
Three follow-up patches for recent strict netlink validation series.
Patch 1 fixes dump handling for genetlink families which validate and parse
messages themselves (e.g. because they need different policies for diferent
commands).
Patch 2 sets bad_attr in extack in one place where this was omitted.
Patch 3 adds new NL_VALIDATE_NESTED flags for strict validation to enable
checking that NLA_F_NESTED value in received messages matches expectations
and includes this flag in NL_VALIDATE_STRICT. This would change userspace
visible behavior but the previous switching to NL_VALIDATE_STRICT for new
code is still only in net-next at the moment.
v2: change error messages to mention NLA_F_NESTED explicitly
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new validation flag NL_VALIDATE_NESTED which adds three consistency
checks of NLA_F_NESTED_FLAG:
- the flag is set on attributes with NLA_NESTED{,_ARRAY} policy
- the flag is not set on attributes with other policies except NLA_UNSPEC
- the flag is set on attribute passed to nla_parse_nested()
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
v2: change error messages to mention NLA_F_NESTED explicitly
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The check that attribute type is within 0...maxtype range in
__nla_validate_parse() sets only error message but not bad_attr in extack.
Set also bad_attr to tell userspace which attribute failed validation.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unlike do requests, dump genetlink requests now perform strict validation
by default even if the genetlink family does not set policy and maxtype
because it does validation and parsing on its own (e.g. because it wants to
allow different message format for different commands). While the null
policy will be ignored, maxtype (which would be zero) is still checked so
that any attribute will fail validation.
The solution is to only call __nla_validate() from genl_family_rcv_msg()
if family->maxtype is set.
Fixes: ef6243acb4 ("genetlink: optionally validate strictly/dumps")
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel says:
====================
mlxsw: Firmware version update
This patchset updates mlxsw to use a new firmware version and adds
support for split into two ports on Spectrum-2 based systems.
Patch #1 updates the firmware version to 13.2000.1122
Patch #2 queries new resources from the firmware.
Patch #3 makes use of these resources in order to support split into two
ports on Spectrum-2 based systems. The need for these resources is
explained by Shalom:
When splitting a port, different local ports need to be mapped on different
systems. For example:
SN3700 (local_ports_in_2x=2):
* Without split:
front panel 1 --> local port 1
front panel 2 --> local port 5
* Split to 2:
front panel 1s0 --> local port 1
front panel 1s1 --> local port 3
front panel 2 --> local port 5
SN3800 (local_ports_in_2x=1):
* Without split:
front panel 1 --> local port 1
front panel 2 --> local port 3
* Split to 2:
front panel 1s0 --> local port 1
front panel 1s1 --> local port 2
front panel 2 --> local port 3
The local_ports_in_{1x, 2x} resources provide the offsets from the base
local ports according to which the new local ports can be calculated.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When splitting a port, different local ports need to be mapped on different
systems. For example:
SN3700 (local_ports_in_2x=2):
* Without split:
front panel 1 --> local port 1
front panel 2 --> local port 5
* Split to 2:
front panel 1s0 --> local port 1
front panel 1s1 --> local port 3
front panel 2 --> local port 5
SN3800 (local_ports_in_2x=1):
* Without split:
front panel 1 --> local port 1
front panel 2 --> local port 3
* Split to 2:
front panel 1s0 --> local port 1
front panel 1s1 --> local port 2
front panel 2 --> local port 3
The local_ports_in_{1x, 2x} resources provide the offsets from the base
local ports according to which the new local ports can be calculated.
Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since the number of local ports in 4x changed between SPC and SPC-2,
firmware expose new resources that the driver can query.
Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The new version supports two features that are required by upcoming
changes in the driver:
* Querying of new resources allowing port split into two ports on
Spectrum-2 systems
* Querying of number of gearboxes on supported systems such as SN3800
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TIPC link can temporarily fall into "half-establish" that only one of
the link endpoints is ESTABLISHED and starts to send traffic, PROTOCOL
messages, whereas the other link endpoint is not up (e.g. immediately
when the endpoint receives ACTIVATE_MSG, the network interface goes
down...).
This is a normal situation and will be settled because the link
endpoint will be eventually brought down after the link tolerance time.
However, the situation will become worse when the second link is
established before the first link endpoint goes down,
For example:
1. Both links <1A-2A>, <1B-2B> down
2. Link endpoint 2A up, but 1A still down (e.g. due to network
disturbance, wrong session, etc.)
3. Link <1B-2B> up
4. Link endpoint 2A down (e.g. due to link tolerance timeout)
5. Node B starts failover onto link <1B-2B>
==> Node A does never start link failover.
When the "half-failover" situation happens, two consequences have been
observed:
a) Peer link/node gets stuck in FAILINGOVER state;
b) Traffic or user messages that peer node is trying to failover onto
the second link can be partially or completely dropped by this node.
The consequence a) was actually solved by commit c140eb166d ("tipc:
fix failover problem"), but that commit didn't cover the b). It's due
to the fact that the tunnel link endpoint has never been prepared for a
failover, so the 'l->drop_point' (and the other data...) is not set
correctly. When a TUNNEL_MSG from peer node arrives on the link,
depending on the inner message's seqno and the current 'l->drop_point'
value, the message can be dropped (- treated as a duplicate message) or
processed.
At this early stage, the traffic messages from peer are likely to be
NAME_DISTRIBUTORs, this means some name table entries will be missed on
the node forever!
The commit resolves the issue by starting the FAILOVER process on this
node as well. Another benefit from this solution is that we ensure the
link will not be re-established until the failover ends.
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
I got an interesting report [0] that after resuming from hibernation
the link has 100Mbps instead of 1Gbps. Reason is that another OS has
been used whilst Linux was hibernated. And this OS speeds down the link
due to WoL. Therefore, when resuming, we shouldn't expect that what
the PHY advertises is what it did when hibernating.
Easiest way to do this is removing state PHY_RESUMING. Instead always
go via PHY_UP that configures PHY advertisement.
[0] https://bugzilla.kernel.org/show_bug.cgi?id=202851
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
When probing the phy device we set sym and asym pause in the "supported"
bitmap (unless the PHY tells us otherwise). However we don't know yet
whether the MAC supports pause. Simply copying phy->supported to
phy->advertising will trigger advertising pause, and that's not
what we want. Therefore add phy_advertise_supported() that copies all
modes but doesn't touch the pause bits.
In phy_support_(a)sym_pause we shouldn't set any bits in the supported
bitmap because we may set a bit the PHY intentionally disabled.
Effective pause support should be the AND-combined PHY and MAC pause
capabilities. If the MAC supports everything, then it's only relevant
what the PHY supports. If MAC supports sym pause only, then we have to
clear the asym bit in phydev->supported.
Copy the pause flags only and don't touch the modes, because a driver
may have intentionally removed a mode from phydev->advertising.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make use of the struct_size() helper instead of an open-coded version
in order to avoid any potential type mistakes, in particular in the
context in which this code is being used.
So, replace code of the following form:
sizeof(*s) + s->nkeys*sizeof(struct tc_u32_key)
with:
struct_size(s, keys, s->nkeys)
This code was detected with the help of Coccinelle.
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Although devlink health report does a nice job on reporting TX
timeout and other NIC errors, unfortunately it requires drivers
to support it but currently only mlx5 has implemented it.
Before other drivers could catch up, it is useful to have a
generic tracepoint to monitor this kind of TX timeout. We have
been suffering TX timeout with different drivers, we plan to
start to monitor it with rasdaemon which just needs a new tracepoint.
Sample output:
ksoftirqd/1-16 [001] ..s2 144.043173: net_dev_xmit_timeout: dev=ens3 driver=e1000 queue=0
Cc: Eran Ben Elisha <eranbe@mellanox.com>
Cc: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit cd9ff4de01 changed the key for IFF_POINTOPOINT devices to
INADDR_ANY but neigh_xmit which is used for MPLS encapsulations was not
updated to use the altered key. The result is that every packet Tx does
a lookup on the gateway address which does not find an entry, a new one
is created only to find the existing one in the table right before the
insert since arp_constructor was updated to reset the primary key. This
is seen in the allocs and destroys counters:
ip -s -4 ntable show | head -10 | grep alloc
which increase for each packet showing the unnecessary overhread.
Fix by having neigh_xmit use __ipv4_neigh_lookup_noref for NEIGH_ARP_TABLE.
Fixes: cd9ff4de01 ("ipv4: Make neigh lookup keys for loopback/point-to-point devices be INADDR_ANY")
Reported-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Tested-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ian and Alan both reported seeing overflows after upgrades to 5.x kernels:
neighbour: arp_cache: neighbor table overflow!
Alan's mpls script helped get to the bottom of this bug. When a new entry
is created the gc_entries counter is bumped in neigh_alloc to check if a
new one is allowed to be created. ___neigh_create then searches for an
existing entry before inserting the just allocated one. If an entry
already exists, the new one is dropped in favor of the existing one. In
this case the cleanup path needs to drop the gc_entries counter. There
is no memory leak, only a counter leak.
Fixes: 58956317c8 ("neighbor: Improve garbage collection")
Reported-by: Ian Kumlien <ian.kumlien@gmail.com>
Reported-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Tested-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>