platform/kernel/linux-starfive.git
5 years agoaio: don't zero entire aio_kiocb aio_get_req()
Jens Axboe [Tue, 4 Dec 2018 16:44:49 +0000 (09:44 -0700)]
aio: don't zero entire aio_kiocb aio_get_req()

It's 192 bytes, fairly substantial. Most items don't need to be cleared,
especially not upfront. Clear the ones we do need to clear, and leave
the other ones for setup when the iocb is prepared and submitted.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoaio: separate out ring reservation from req allocation
Christoph Hellwig [Mon, 19 Nov 2018 22:57:42 +0000 (15:57 -0700)]
aio: separate out ring reservation from req allocation

This is in preparation for certain types of IO not needing a ring
reserveration.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoaio: use assigned completion handler
Jens Axboe [Tue, 6 Nov 2018 21:27:13 +0000 (14:27 -0700)]
aio: use assigned completion handler

We know this is a read/write request, but in preparation for
having different kinds of those, ensure that we call the assigned
handler instead of assuming it's aio_complete_rq().

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoMerge branch 'for-4.21/block' into for-4.21/aio
Jens Axboe [Tue, 18 Dec 2018 15:29:53 +0000 (08:29 -0700)]
Merge branch 'for-4.21/block' into for-4.21/aio

* for-4.21/block: (351 commits)
  blk-mq: enable IO poll if .nr_queues of type poll > 0
  blk-mq: change blk_mq_queue_busy() to blk_mq_queue_inflight()
  blk-mq: skip zero-queue maps in blk_mq_map_swqueue
  block: fix blk-iolatency accounting underflow
  blk-mq: fix dispatch from sw queue
  block: mq-deadline: Fix write completion handling
  nvme-pci: don't share queue maps
  blk-mq: only dispatch to non-defauly queue maps if they have queues
  blk-mq: export hctx->type in debugfs instead of sysfs
  blk-mq: fix allocation for queue mapping table
  blk-wbt: export internal state via debugfs
  blk-mq-debugfs: support rq_qos
  block: update sysfs documentation
  block: loop: check error using IS_ERR instead of IS_ERR_OR_NULL in loop_add()
  aoe: add __exit annotation
  block: clear REQ_HIPRI if polling is not supported
  blk-mq: replace and kill blk_mq_request_issue_directly
  blk-mq: issue directly with bypass 'false' in blk_mq_sched_insert_requests
  blk-mq: refactor the code of issue request directly
  block: remove the bio_integrity_advance export
  ...

5 years agoblk-mq: enable IO poll if .nr_queues of type poll > 0
Ming Lei [Tue, 18 Dec 2018 04:15:29 +0000 (12:15 +0800)]
blk-mq: enable IO poll if .nr_queues of type poll > 0

The queue mapping of type poll only exists when set->map[HCTX_TYPE_POLL].nr_queues
is bigger than zero, so enhance the constraint by checking .nr_queues of type poll
before enabling IO poll.

Otherwise IO race & timeout can be observed when running block/007.

Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: change blk_mq_queue_busy() to blk_mq_queue_inflight()
Jens Axboe [Tue, 18 Dec 2018 04:11:17 +0000 (21:11 -0700)]
blk-mq: change blk_mq_queue_busy() to blk_mq_queue_inflight()

There's a single user of this function, dm, and dm just wants
to check if IO is inflight, not that it's just allocated.

This fixes a hang with srp/002 in blktests with dm, where it tries
to suspend but waits for inflight IO to finish first. As it checks
for just allocated requests, this fails.

Tested-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoima: cleanup the match_token policy code
Mimi Zohar [Tue, 18 Dec 2018 00:14:49 +0000 (19:14 -0500)]
ima: cleanup the match_token policy code

Start the policy_tokens and the associated enumeration from zero,
simplifying the pt macro.

Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agosecurity: don't use a negative Opt_err token index
Linus Torvalds [Mon, 17 Dec 2018 19:39:57 +0000 (11:39 -0800)]
security: don't use a negative Opt_err token index

The code uses a bitmap to check for duplicate tokens during parsing, and
that doesn't work at all for the negative Opt_err token case.

There is absolutely no reason to make Opt_err be negative, and in fact
it only confuses things, since some of the affected functions actually
return a positive Opt_xyz enum _or_ a regular negative error code (eg
-EINVAL), and using -1 for Opt_err makes no sense.

There are similar problems in ima_policy.c and key encryption, but they
don't have the immediate bug wrt bitmap handing, and ima_policy.c in
particular needs a different patch to make the enum values match the
token array index.  Mimi is sending that separately.

Reported-by: syzbot+a22e0dc07567662c50bc@syzkaller.appspotmail.com
Reported-by: Eric Biggers <ebiggers@kernel.org>
Fixes: 5208cc83423d ("keys, trusted: fix: *do not* allow duplicate key options")
Fixes: 00d60fd3b932 ("KEYS: Provide keyctls to drive the new key type ops for asymmetric keys [ver #2]")
Cc: James Morris James Morris <jmorris@namei.org>
Cc: Mimi Zohar <zohar@linux.vnet.ibm.com>
Cc: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Cc: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoblk-mq: skip zero-queue maps in blk_mq_map_swqueue
Ming Lei [Mon, 17 Dec 2018 17:28:56 +0000 (01:28 +0800)]
blk-mq: skip zero-queue maps in blk_mq_map_swqueue

From 7e849dd9cf37 ("nvme-pci: don't share queue maps"), the mapping
table won't be initialized actually if map->nr_queues is zero, so
we can't use blk_mq_map_queue_type() to retrieve hctx any more.

This way still may cause broken mapping, fix it by skipping zero-queues
maps in blk_mq_map_swqueue().

Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: fix blk-iolatency accounting underflow
Dennis Zhou [Mon, 17 Dec 2018 16:03:51 +0000 (11:03 -0500)]
block: fix blk-iolatency accounting underflow

The blk-iolatency controller measures the time from rq_qos_throttle() to
rq_qos_done_bio() and attributes this time to the first bio that needs
to create the request. This means if a bio is plug-mergeable or
bio-mergeable, it gets to bypass the blk-iolatency controller.

The recent series [1], to tag all bios w/ blkgs undermined how iolatency
was determining which bios it was charging and should process in
rq_qos_done_bio(). Because all bios are being tagged, this caused the
atomic_t for the struct rq_wait inflight count to underflow and result
in a stall.

This patch adds a new flag BIO_TRACKED to let controllers know that a
bio is going through the rq_qos path. blk-iolatency now checks if this
flag is set to see if it should process the bio in rq_qos_done_bio().

Overloading BLK_QUEUE_ENTERED works, but makes the flag rules confusing.
BIO_THROTTLED was another candidate, but the flag is set for all bios
that have gone through blk-throttle code. Overloading a flag comes with
the burden of making sure that when either implementation changes, a
change in setting rules for one doesn't cause a bug in the other. So
here, we unfortunately opt for adding a new flag.

[1] https://lore.kernel.org/lkml/20181205171039.73066-1-dennis@kernel.org/

Fixes: 5cdf2e3fea5e ("blkcg: associate blkg when associating a device")
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Cc: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: fix dispatch from sw queue
Ming Lei [Mon, 17 Dec 2018 15:44:05 +0000 (08:44 -0700)]
blk-mq: fix dispatch from sw queue

When a request is added to rq list of sw queue(ctx), the rq may be from
a different type of hctx, especially after multi queue mapping is
introduced.

So when dispach request from sw queue via blk_mq_flush_busy_ctxs() or
blk_mq_dequeue_from_ctx(), one request belonging to other queue type of
hctx can be dispatched to current hctx in case that read queue or poll
queue is enabled.

This patch fixes this issue by introducing per-queue-type list.

Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Changed by me to not use separately cacheline aligned lists, just
place them all in the same cacheline where we had just the one list
and lock before.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: mq-deadline: Fix write completion handling
Damien Le Moal [Mon, 17 Dec 2018 06:14:05 +0000 (15:14 +0900)]
block: mq-deadline: Fix write completion handling

For a zoned block device using mq-deadline, if a write request for a
zone is received while another write was already dispatched for the same
zone, dd_dispatch_request() will return NULL and the newly inserted
write request is kept in the scheduler queue waiting for the ongoing
zone write to complete. With this behavior, when no other request has
been dispatched, rq_list in blk_mq_sched_dispatch_requests() is empty
and blk_mq_sched_mark_restart_hctx() not called. This in turn leads to
__blk_mq_free_request() call of blk_mq_sched_restart() to not run the
queue when the already dispatched write request completes. The newly
dispatched request stays stuck in the scheduler queue until eventually
another request is submitted.

This problem does not affect SCSI disk as the SCSI stack handles queue
restart on request completion. However, this problem is can be triggered
the nullblk driver with zoned mode enabled.

Fix this by always requesting a queue restart in dd_dispatch_request()
if no request was dispatched while WRITE requests are queued.

Fixes: 5700f69178e9 ("mq-deadline: Introduce zone locking support")
Cc: <stable@vger.kernel.org>
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Add missing export of blk_mq_sched_restart()

Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonvme-pci: don't share queue maps
Christoph Hellwig [Mon, 17 Dec 2018 11:16:27 +0000 (12:16 +0100)]
nvme-pci: don't share queue maps

Now that the block layer checks if a queue map has any queues inside
it there is no more reason to duplicate the maps for the non-default
types.

Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: only dispatch to non-defauly queue maps if they have queues
Christoph Hellwig [Mon, 17 Dec 2018 11:16:26 +0000 (12:16 +0100)]
blk-mq: only dispatch to non-defauly queue maps if they have queues

We should check if a given queue map actually has queues enabled before
dispatching to it.  This allows drivers to not initialize optional but
not used map types, which subsequently will allow fixing problems with
queue map rebuilds for that case.

Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: export hctx->type in debugfs instead of sysfs
Ming Lei [Mon, 17 Dec 2018 10:42:48 +0000 (18:42 +0800)]
blk-mq: export hctx->type in debugfs instead of sysfs

Now we only export hctx->type via sysfs, and there isn't such info
in hctx entry under debugfs. We often use debugfs only to diagnose
queue mapping issue, so add the support in debugfs.

Queue mapping becomes a bit more complicated after multiple queue
mapping is supported, we may write blktest to verify if queue mapping
is valid based on blk-mq-debugfs.

Given not necessary to export hctx->type twice, so remove the export
from sysfs.

Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: fix allocation for queue mapping table
Ming Lei [Mon, 17 Dec 2018 10:42:45 +0000 (18:42 +0800)]
blk-mq: fix allocation for queue mapping table

Type of each element in queue mapping table is 'unsigned int,
intead of 'struct blk_mq_queue_map)', so fix it.

Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Mike Snitzer <snitzer@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-wbt: export internal state via debugfs
Ming Lei [Mon, 17 Dec 2018 01:46:01 +0000 (09:46 +0800)]
blk-wbt: export internal state via debugfs

This information is helpful to either investigate issues, or understand
wbt's internal behaviour.

Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Omar Sandoval <osandov@fb.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq-debugfs: support rq_qos
Ming Lei [Mon, 17 Dec 2018 01:46:00 +0000 (09:46 +0800)]
blk-mq-debugfs: support rq_qos

blk-mq-debugfs has been proved as very helpful for debug some
tough issues, such as IO hang.

We have seen blk-wbt related IO hang several times, even inside
Red Hat BZ, there is such report not sovled yet, so this patch
adds support debugfs on rq_qos.

Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Omar Sandoval <osandov@fb.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: update sysfs documentation
Damien Le Moal [Fri, 30 Nov 2018 05:36:24 +0000 (14:36 +0900)]
block: update sysfs documentation

Add the description of the zoned, nr_zones and chunk_sectors sysfs queue
attributes to Documentation/block/queue-sysfs.txt. The description of
the zoned and chunk_sector attributes are mostly copied from
ABI/testing/sysfs-block (added a typo fix). While at it, also fix a
typo in the description of the io_poll_delay attribute.

nr_zones description is also added to ABI/testing/sysfs-block and
contact email address updated for the zoned attribute.

Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoLinux 4.20-rc7
Linus Torvalds [Sun, 16 Dec 2018 23:46:55 +0000 (15:46 -0800)]
Linux 4.20-rc7

5 years agoblock: loop: check error using IS_ERR instead of IS_ERR_OR_NULL in loop_add()
Chengguang Xu [Sun, 16 Dec 2018 09:35:00 +0000 (17:35 +0800)]
block: loop: check error using IS_ERR instead of IS_ERR_OR_NULL in loop_add()

blk_mq_init_queue() will not return NULL pointer to its caller,
so it's better to replace IS_ERR_OR_NULL using IS_ERR in loop_add().

If in the future things change to check NULL pointer inside loop_add(),
we should return -ENOMEM as return code instead of PTR_ERR(NULL).

Signed-off-by: Chengguang Xu <cgxu519@gmx.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoaoe: add __exit annotation
Chengguang Xu [Sun, 16 Dec 2018 06:08:18 +0000 (14:08 +0800)]
aoe: add __exit annotation

Add __exit annotation to cleanup helper which
is only called once in the module.

Signed-off-by: Chengguang Xu <cgxu519@gmx.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: clear REQ_HIPRI if polling is not supported
Christoph Hellwig [Fri, 14 Dec 2018 16:21:22 +0000 (17:21 +0100)]
block: clear REQ_HIPRI if polling is not supported

This prevents a HIPRI bio from being submitted through a stacking
driver that does not support polling and thus won't poll for I/O
completion.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: replace and kill blk_mq_request_issue_directly
Jianchao Wang [Fri, 14 Dec 2018 01:28:20 +0000 (09:28 +0800)]
blk-mq: replace and kill blk_mq_request_issue_directly

Replace blk_mq_request_issue_directly with blk_mq_try_issue_directly
in blk_insert_cloned_request and kill it as nobody uses it any more.

Signed-off-by: Jianchao Wang <jianchao.w.wang@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: issue directly with bypass 'false' in blk_mq_sched_insert_requests
Jianchao Wang [Fri, 14 Dec 2018 01:28:19 +0000 (09:28 +0800)]
blk-mq: issue directly with bypass 'false' in blk_mq_sched_insert_requests

It is not necessary to issue request directly with bypass 'true'
in blk_mq_sched_insert_requests and handle the non-issued requests
itself. Just set bypass to 'false' and let blk_mq_try_issue_directly
handle them totally. Remove the blk_rq_can_direct_dispatch check,
because blk_mq_try_issue_directly can handle it well.If request is
direct-issued unsuccessfully, insert the reset.

Signed-off-by: Jianchao Wang <jianchao.w.wang@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: refactor the code of issue request directly
Jianchao Wang [Fri, 14 Dec 2018 01:28:18 +0000 (09:28 +0800)]
blk-mq: refactor the code of issue request directly

Merge blk_mq_try_issue_directly and __blk_mq_try_issue_directly
into one interface to unify the interfaces to issue requests
directly. The merged interface takes over the requests totally,
it could insert, end or do nothing based on the return value of
.queue_rq and 'bypass' parameter. Then caller needn't any other
handling any more and then code could be cleaned up.

And also the commit c616cbee ( blk-mq: punt failed direct issue
to dispatch list ) always inserts requests to hctx dispatch list
whenever get a BLK_STS_RESOURCE or BLK_STS_DEV_RESOURCE, this is
overkill and will harm the merging. We just need to do that for
the requests that has been through .queue_rq. This patch also
could fix this.

Signed-off-by: Jianchao Wang <jianchao.w.wang@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: remove the bio_integrity_advance export
Christoph Hellwig [Thu, 13 Dec 2018 20:32:14 +0000 (21:32 +0100)]
block: remove the bio_integrity_advance export

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: remove the bioset_integrity_free export
Christoph Hellwig [Thu, 13 Dec 2018 20:32:13 +0000 (21:32 +0100)]
block: remove the bioset_integrity_free export

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 14 Dec 2018 23:35:30 +0000 (15:35 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "11 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  scripts/spdxcheck.py: always open files in binary mode
  checkstack.pl: fix for aarch64
  userfaultfd: check VM_MAYWRITE was set after verifying the uffd is registered
  fs/iomap.c: get/put the page in iomap_page_create/release()
  hugetlbfs: call VM_BUG_ON_PAGE earlier in free_huge_page()
  memblock: annotate memblock_is_reserved() with __init_memblock
  psi: fix reference to kernel commandline enable
  arch/sh/include/asm/io.h: provide prototypes for PCI I/O mapping in asm/io.h
  mm/sparse: add common helper to mark all memblocks present
  mm: introduce common STRUCT_PAGE_MAX_SHIFT define
  alpha: fix hang caused by the bootmem removal

5 years agoscripts/spdxcheck.py: always open files in binary mode
Thierry Reding [Fri, 14 Dec 2018 22:17:24 +0000 (14:17 -0800)]
scripts/spdxcheck.py: always open files in binary mode

The spdxcheck script currently falls over when confronted with a binary
file (such as Documentation/logo.gif).  To avoid that, always open files
in binary mode and decode line-by-line, ignoring encoding errors.

One tricky case is when piping data into the script and reading it from
standard input.  By default, standard input will be opened in text mode,
so we need to reopen it in binary mode.

The breakage only happens with python3 and results in a
UnicodeDecodeError (according to Uwe).

Link: http://lkml.kernel.org/r/20181212131210.28024-1-thierry.reding@gmail.com
Fixes: 6f4d29df66ac ("scripts/spdxcheck.py: make python3 compliant")
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Jeremy Cline <jcline@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Joe Perches <joe@perches.com>
Cc: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agocheckstack.pl: fix for aarch64
Qian Cai [Fri, 14 Dec 2018 22:17:20 +0000 (14:17 -0800)]
checkstack.pl: fix for aarch64

There is actually a space after "sp," like this,

    ffff2000080813c8:       a9bb7bfd        stp     x29, x30, [sp, #-80]!

Right now, checkstack.pl isn't able to print anything on aarch64,
because it won't be able to match the stating objdump line of a function
due to this missing space.  Hence, it displays every stack as zero-size.

After this patch, checkpatch.pl is able to match the start of a
function's objdump, and is then able to calculate each function's stack
correctly.

Link: http://lkml.kernel.org/r/20181207195843.38528-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agouserfaultfd: check VM_MAYWRITE was set after verifying the uffd is registered
Andrea Arcangeli [Fri, 14 Dec 2018 22:17:17 +0000 (14:17 -0800)]
userfaultfd: check VM_MAYWRITE was set after verifying the uffd is registered

Calling UFFDIO_UNREGISTER on virtual ranges not yet registered in uffd
could trigger an harmless false positive WARN_ON.  Check the vma is
already registered before checking VM_MAYWRITE to shut off the false
positive warning.

Link: http://lkml.kernel.org/r/20181206212028.18726-2-aarcange@redhat.com
Cc: <stable@vger.kernel.org>
Fixes: 29ec90660d68 ("userfaultfd: shmem/hugetlbfs: only allow to register VM_MAYWRITE vmas")
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: syzbot+06c7092e7d71218a2c16@syzkaller.appspotmail.com
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agofs/iomap.c: get/put the page in iomap_page_create/release()
Piotr Jaroszynski [Fri, 14 Dec 2018 22:17:14 +0000 (14:17 -0800)]
fs/iomap.c: get/put the page in iomap_page_create/release()

migrate_page_move_mapping() expects pages with private data set to have
a page_count elevated by 1.  This is what used to happen for xfs through
the buffer_heads code before the switch to iomap in commit 82cb14175e7d
("xfs: add support for sub-pagesize writeback without buffer_heads").
Not having the count elevated causes move_pages() to fail on memory
mapped files coming from xfs.

Make iomap compatible with the migrate_page_move_mapping() assumption by
elevating the page count as part of iomap_page_create() and lowering it
in iomap_page_release().

It causes the move_pages() syscall to misbehave on memory mapped files
from xfs.  It does not not move any pages, which I suppose is "just" a
perf issue, but it also ends up returning a positive number which is out
of spec for the syscall.  Talking to Michal Hocko, it sounds like
returning positive numbers might be a necessary update to move_pages()
anyway though
(https://lkml.kernel.org/r/20181116114955.GJ14706@dhcp22.suse.cz).

I only hit this in tests that verify that move_pages() actually moved
the pages.  The test also got confused by the positive return from
move_pages() (it got treated as a success as positive numbers were not
expected and not handled) making it a bit harder to track down what's
going on.

Link: http://lkml.kernel.org/r/20181115184140.1388751-1-pjaroszynski@nvidia.com
Fixes: 82cb14175e7d ("xfs: add support for sub-pagesize writeback without buffer_heads")
Signed-off-by: Piotr Jaroszynski <pjaroszynski@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Brian Foster <bfoster@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agohugetlbfs: call VM_BUG_ON_PAGE earlier in free_huge_page()
Yongkai Wu [Fri, 14 Dec 2018 22:17:10 +0000 (14:17 -0800)]
hugetlbfs: call VM_BUG_ON_PAGE earlier in free_huge_page()

A stack trace was triggered by VM_BUG_ON_PAGE(page_mapcount(page), page)
in free_huge_page().  Unfortunately, the page->mapping field was set to
NULL before this test.  This made it more difficult to determine the
root cause of the problem.

Move the VM_BUG_ON_PAGE tests earlier in the function so that if they do
trigger more information is present in the page struct.

Link: http://lkml.kernel.org/r/1543491843-23438-1-git-send-email-nic_w@163.com
Signed-off-by: Yongkai Wu <nic_w@163.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomemblock: annotate memblock_is_reserved() with __init_memblock
Yueyi Li [Fri, 14 Dec 2018 22:17:06 +0000 (14:17 -0800)]
memblock: annotate memblock_is_reserved() with __init_memblock

Found warning:

  WARNING: EXPORT symbol "gsi_write_channel_scratch" [vmlinux] version generation failed, symbol will not be versioned.
  WARNING: vmlinux.o(.text+0x1e0a0): Section mismatch in reference from the function valid_phys_addr_range() to the function .init.text:memblock_is_reserved()
  The function valid_phys_addr_range() references
  the function __init memblock_is_reserved().
  This is often because valid_phys_addr_range lacks a __init
  annotation or the annotation of memblock_is_reserved is wrong.

Use __init_memblock instead of __init.

Link: http://lkml.kernel.org/r/BLUPR13MB02893411BF12EACB61888E80DFAE0@BLUPR13MB0289.namprd13.prod.outlook.com
Signed-off-by: Yueyi Li <liyueyi@live.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agopsi: fix reference to kernel commandline enable
Baruch Siach [Fri, 14 Dec 2018 22:17:03 +0000 (14:17 -0800)]
psi: fix reference to kernel commandline enable

The kernel commandline parameter named in CONFIG_PSI_DEFAULT_DISABLED
help text contradicts the documentation in kernel-parameters.txt, and
the code.  Fix that.

Link: http://lkml.kernel.org/r/20181203213416.GA12627@cmpxchg.org
Fixes: e0c274472d ("psi: make disabling/enabling easier for vendor kernels")
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoarch/sh/include/asm/io.h: provide prototypes for PCI I/O mapping in asm/io.h
Mark Brown [Fri, 14 Dec 2018 22:17:00 +0000 (14:17 -0800)]
arch/sh/include/asm/io.h: provide prototypes for PCI I/O mapping in asm/io.h

Most architectures provide prototypes for the PCI I/O mapping operations
when asm/io.h is included but SH doesn't currently do that, leading to
for example warnings in sound/pci/hda/patch_ca0132.c when pci_iomap() is
used on current -next.  Make SH more consistent with other architectures
by including asm-generic/pci_iomap.h in asm/io.h.

Link: http://lkml.kernel.org/r/20181106175142.27988-1-broonie@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Reported-by: kbuild test robot <lkp@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Rich Felker <dalias@libc.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/sparse: add common helper to mark all memblocks present
Logan Gunthorpe [Fri, 14 Dec 2018 22:16:57 +0000 (14:16 -0800)]
mm/sparse: add common helper to mark all memblocks present

Presently the arches arm64, arm and sh have a function which loops
through each memblock and calls memory present.  riscv will require a
similar function.

Introduce a common memblocks_present() function that can be used by all
the arches.  Subsequent patches will cleanup the arches that make use of
this.

Link: http://lkml.kernel.org/r/20181107205433.3875-3-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm: introduce common STRUCT_PAGE_MAX_SHIFT define
Logan Gunthorpe [Fri, 14 Dec 2018 22:16:53 +0000 (14:16 -0800)]
mm: introduce common STRUCT_PAGE_MAX_SHIFT define

This define is used by arm64 to calculate the size of the vmemmap
region.  It is defined as the log2 of the upper bound on the size of a
struct page.

We move it into mm_types.h so it can be defined properly instead of set
and checked with a build bug.  This also allows us to use the same
define for riscv.

Link: http://lkml.kernel.org/r/20181107205433.3875-2-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoalpha: fix hang caused by the bootmem removal
Mike Rapoport [Fri, 14 Dec 2018 22:16:50 +0000 (14:16 -0800)]
alpha: fix hang caused by the bootmem removal

The conversion of alpha to memblock as the early memory manager caused
boot to hang as described at [1].

The issue is caused because for CONFIG_DISCTONTIGMEM=y case,
memblock_add() is called using memory start PFN that had been rounded
down to the nearest 8Mb and it caused memblock to see more memory that
is actually present in the system.

Besides, memblock allocates memory from high addresses while bootmem was
using low memory, which broke the assumption that early allocations are
always accessible by the hardware.

This patch ensures that memblock_add() is using the correct PFN for the
memory start and forces memblock to use bottom-up allocations.

[1] https://lkml.org/lkml/2018/11/22/1032

Link: http://lkml.kernel.org/r/1543233216-25833-1-git-send-email-rppt@linux.ibm.com
Reported-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Tested-by: Meelis Roos <mroos@linux.ee>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoMerge tag 'for-linus-20181214' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 14 Dec 2018 20:18:30 +0000 (12:18 -0800)]
Merge tag 'for-linus-20181214' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "Three small fixes for this week. contains:

   - spectre indexing fix for aio (Jeff)

   - fix for the previous zeroing bio fix, we don't need it for user
     mapped pages, and in fact it breaks some applications if we do
     (Keith)

   - allocation failure fix for null_blk with zoned (Shin'ichiro)"

* tag 'for-linus-20181214' of git://git.kernel.dk/linux-block:
  block: Fix null_blk_zoned creation failure with small number of zones
  aio: fix spectre gadget in lookup_ioctx
  block/bio: Do not zero user pages

5 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 14 Dec 2018 20:14:41 +0000 (12:14 -0800)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fix from Stephen Boyd:
 "One fix for the qcom QCS404 clk driver that was merged for this
  release.

  It specified the wrong parent for a PLL so a part of the clk tree
  wasn't rooted correctly. This fixes it by using the right name"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: qcom: qcs404: Fix gpll0_out_main parent

5 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 14 Dec 2018 17:36:41 +0000 (09:36 -0800)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fix from Catalin Marinas:
 "Invalidate the caches before clearing the DMA buffer via the
  non-cacheable alias in the FORCE_CONTIGUOUS case"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: dma-mapping: Fix FORCE_CONTIGUOUS buffer clearing

5 years agoMerge tag 'powerpc-4.20-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 14 Dec 2018 17:33:34 +0000 (09:33 -0800)]
Merge tag 'powerpc-4.20-4' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "One notable fix for our change to split pt_regs between user/kernel,
  we forgot to update BPF to use the user-visible type which was an ABI
  break for BPF programs.

  A slightly ugly but minimal fix to do_syscall_trace_enter() so that we
  use tracehook_report_syscall_entry() properly. We'll rework the code
  in next to avoid the empty if body.

  Seven commits fixing bugs in the new papr_scm (Storage Class Memory)
  driver. The driver was finally able to be tested on the other
  hypervisor which exposed several bugs. The fixes are all fairly
  minimal at least.

  Fix a crash in our MSI code if an MSI-capable device is plugged into a
  non-MSI capable PHB, only seen on older hardware (MPC8378).

  Fix our legacy serial code to look for "stdout-path" since the device
  trees were updated to use that instead of "linux,stdout-path".

  A change to the COFF zImage code to fix booting old powermacs.

  A couple of minor build fixes.

  Thanks to: Benjamin Herrenschmidt, Daniel Axtens, Dmitry V. Levin,
  Elvira Khabirova, Oliver O'Halloran, Paul Mackerras, Radu Rendec, Rob
  Herring, Sandipan Das"

* tag 'powerpc-4.20-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/ptrace: replace ptrace_report_syscall() with a tracehook call
  powerpc/mm: Fallback to RAM if the altmap is unusable
  powerpc/papr_scm: Use ibm,unit-guid as the iset cookie
  powerpc/papr_scm: Fix DIMM device registration race
  powerpc/papr_scm: Remove endian conversions
  powerpc/papr_scm: Update DT properties
  powerpc/papr_scm: Fix resource end address
  powerpc/papr_scm: Use depend instead of select
  powerpc/bpf: Fix broken uapi for BPF_PROG_TYPE_PERF_EVENT
  powerpc/boot: Fix build failures with -j 1
  powerpc: Look for "stdout-path" when setting up legacy consoles
  powerpc/msi: Fix NULL pointer access in teardown code
  powerpc/mm: Fix linux page tables build with some configs
  powerpc: Fix COFF zImage booting on old powermacs

5 years agoMerge tag 'ceph-for-4.20-rc7' of https://github.com/ceph/ceph-client
Linus Torvalds [Fri, 14 Dec 2018 17:22:14 +0000 (09:22 -0800)]
Merge tag 'ceph-for-4.20-rc7' of https://github.com/ceph/ceph-client

Pull ceph fix from Ilya Dryomov:
 "Luis discovered a problem with the new copyfrom offload on the server
  side. Disable it for now"

* tag 'ceph-for-4.20-rc7' of https://github.com/ceph/ceph-client:
  ceph: make 'nocopyfrom' a default mount option

5 years agoMerge tag 'pinctrl-v4.20-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Fri, 14 Dec 2018 17:17:17 +0000 (09:17 -0800)]
Merge tag 'pinctrl-v4.20-3' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull pin control fixes from Linus Walleij:
 "Three pin control fixes for the v4.20 series. Just odd drivers, so
  nothing particularly interesting:

   - Set the tile property on Qualcomm SDM60.

   - Fix up enable register calculation for the Meson

   - Fix an IRQ offset on the Sunxi (Allwinner)"

* tag 'pinctrl-v4.20-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: sunxi: a83t: Fix IRQ offset typo for PH11
  pinctrl: meson: fix pull enable register calculation
  pinctrl: sdm660: Set tile property for pingroups

5 years agoMerge tag 'drm-fixes-2018-12-14' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 14 Dec 2018 17:12:02 +0000 (09:12 -0800)]
Merge tag 'drm-fixes-2018-12-14' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "While I hoped things would calm down, the world hasn't joined with me,
  but it's a few things scattered over a wide area. The i915 workarounds
  regression fix is probably the largest, the rest are more usual sized.
  We also get some new AMD PCI IDs.

  There is also a patch in here to MAINTAINERS to added Daniel as an
  official DRM toplevel co-maintainer, he's decided he wants to step up
  and share the glory, and he'll likely process next weeks fixes while
  I'm away on holidays.

  Summary:

  amdgpu:

   - some new PCI IDs

   - fixed firmware image updates

   - power management fixes

   - locking warning fix

  nouveau:

   - framebuffer flushing fix

   - memory leak fix

   - tegra device init regression fix

  vmwgfx:

   - OOM kernel memory fix

   - excess return in function fix

  i915:

   - the biggest fix is a regression fix where workarounds weren't
     getting reapplied after a gpu hang causing further crashing, this
     fixes the workaround application to make it happen again

   - GPU hang fixes for Braswell and some GEN3 GPUs

   - GVT fix for broadwell tiling

  rockchip:

   - revert to fix a regression causing a WARN on shutdown

  mediatek:

   - avoid crash attaching to non-existant bridges"

* tag 'drm-fixes-2018-12-14' of git://anongit.freedesktop.org/drm/drm: (23 commits)
  drm/vmwgfx: Protect from excessive execbuf kernel memory allocations v3
  MAINTAINERS: Daniel for drm co-maintainer
  drm/amdgpu: drop fclk/gfxclk ratio setting
  drm/vmwgfx: remove redundant return ret statement
  drm/i915: Flush GPU relocs harder for gen3
  drm/i915: Allocate a common scratch page
  drm/i915/execlists: Apply a full mb before execution for Braswell
  drm/nouveau/kms: Fix memory leak in nv50_mstm_del()
  drm/nouveau/kms/nv50-: also flush fb writes when rewinding push buffer
  drm/amdgpu: Fix DEBUG_LOCKS_WARN_ON(depth <= 0) in amdgpu_ctx.lock
  Revert "drm/rockchip: Allow driver to be shutdown on reboot/kexec"
  drm/nouveau/drm/nouveau: tegra: Call nouveau_drm_device_init()
  drm/amdgpu/powerplay: Apply avfs cks-off voltages on VI
  drm/amdgpu: update SMC firmware image for polaris10 variants
  drm/amdkfd: add new vega20 pci id
  drm/amdkfd: add new vega10 pci ids
  drm/amdgpu: add some additional vega20 pci ids
  drm/amdgpu: add some additional vega10 pci ids
  drm/amdgpu: update smu firmware images for VI variants (v2)
  drm/i915: Introduce per-engine workarounds
  ...

5 years agoblock: remove the unused bio_set_pages_dirty and bio_check_pages_dirty exports
Christoph Hellwig [Thu, 13 Dec 2018 20:32:12 +0000 (21:32 +0100)]
block: remove the unused bio_set_pages_dirty and bio_check_pages_dirty exports

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: remove the unused bio_iov_iter_get_pages export
Christoph Hellwig [Thu, 13 Dec 2018 20:32:11 +0000 (21:32 +0100)]
block: remove the unused bio_iov_iter_get_pages export

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: remove the blk_recount_segments export
Christoph Hellwig [Thu, 13 Dec 2018 20:32:10 +0000 (21:32 +0100)]
block: remove the blk_recount_segments export

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: remove the bio_phys_segments export
Christoph Hellwig [Thu, 13 Dec 2018 20:32:09 +0000 (21:32 +0100)]
block: remove the bio_phys_segments export

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoMerge tag 'xarray-4.20-rc7' of git://git.infradead.org/users/willy/linux-dax
Linus Torvalds [Fri, 14 Dec 2018 00:35:58 +0000 (16:35 -0800)]
Merge tag 'xarray-4.20-rc7' of git://git.infradead.org/users/willy/linux-dax

Pull XArray fixes from Matthew Wilcox:
 "Two bugfixes, each with test-suite updates, two improvements to the
  test-suite without associated bugs, and one patch adding a missing
  API"

* tag 'xarray-4.20-rc7' of git://git.infradead.org/users/willy/linux-dax:
  XArray: Fix xa_alloc when id exceeds max
  XArray tests: Check iterating over multiorder entries
  XArray tests: Handle larger indices more elegantly
  XArray: Add xa_cmpxchg_irq and xa_cmpxchg_bh
  radix tree: Don't return retry entries from lookup

5 years agonvme: fix kernel paging oops
Sagi Grimberg [Thu, 13 Dec 2018 20:34:07 +0000 (12:34 -0800)]
nvme: fix kernel paging oops

free the controller discard_page correctly.

Fixes: cb5b7262b011 ("nvme: provide fallback for discard alloc failure")
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoMerge tag 'linux-kselftest-4.20-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 13 Dec 2018 20:57:21 +0000 (12:57 -0800)]
Merge tag 'linux-kselftest-4.20-rc7' of git://git./linux/kernel/git/shuah/linux-kselftest

Pull Kselftest fix from Shuah Khan:
 "A single fix for a seccomp test from Kees Cook."

* tag 'linux-kselftest-4.20-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  selftests/seccomp: Remove SIGSTOP si_pid check

5 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/evalenti/linux...
Linus Torvalds [Thu, 13 Dec 2018 20:02:41 +0000 (12:02 -0800)]
Merge branch 'fixes' of git://git./linux/kernel/git/evalenti/linux-soc-thermal

Pull thermal fixes from Eduardo Valentin:
 "Fixes for STM and HISI thermal drivers"

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/evalenti/linux-soc-thermal:
  thermal: stm32: Fix stm_thermal_read_factory_settings
  thermal: stm32: read factory settings inside stm_thermal_prepare
  thermal/drivers/hisi: Fix number of sensors on hi3660
  thermal/drivers/hisi: Fix wrong platform_get_irq_byname()

5 years agoMerge branch 'nvme-4.21' of git://git.infradead.org/nvme into for-4.21/block
Jens Axboe [Thu, 13 Dec 2018 19:59:37 +0000 (12:59 -0700)]
Merge branch 'nvme-4.21' of git://git.infradead.org/nvme into for-4.21/block

Pull NVMe updates from Christoph:

"Here is the second large chunk of nvme updates for 4.21:

 - host and target support for NVMe over TCP (Sagi Grimberg,
Roy Shterman, Solganik Alexander)
 - error log page support in target (Chaitanya Kulkarni)

plus small fixes and improvements from Jens Axboe and Chengguang Xu."

* 'nvme-4.21' of git://git.infradead.org/nvme: (33 commits)
  nvme-rdma: support separate queue maps for read and write
  nvme-tcp: support separate queue maps for read and write
  nvme-fabrics: allow user to set nr_write_queues for separate queue maps
  nvme-fabrics: add missing nvmf_ctrl_options documentation
  blk-mq-rdma: pass in queue map to blk_mq_rdma_map_queues
  nvmet: update smart log with num err log entries
  nvmet: add error log page cmd handler
  nvmet: add error log support for file backend
  nvmet: add error log support for bdev backend
  nvmet: add error log support for admin-cmd
  nvmet: add error log support for rdma backend
  nvmet: add error log support for fabrics-cmd
  nvmet: add error log support in the core
  nvmet: add interface to update error-log page
  nvmet: add error-log definitions
  nvme: add error log page slot definition
  nvme: remove nvme_common command cdw10 array
  nvmet: remove unused variable
  nvme: provide fallback for discard alloc failure
  nvme: add __exit annotation
  ...

5 years agoMerge branch 'vmwgfx-fixes-4.20' of git://people.freedesktop.org/~thomash/linux into...
Dave Airlie [Thu, 13 Dec 2018 19:19:25 +0000 (05:19 +1000)]
Merge branch 'vmwgfx-fixes-4.20' of git://people.freedesktop.org/~thomash/linux into drm-fixes

One regression fix for avoiding kernel OOM, one cleanup return fix.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Hellstrom <thellstrom@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181213122815.10581-1-thellstrom@vmware.com
5 years agoXArray: Fix xa_alloc when id exceeds max
Matthew Wilcox [Thu, 13 Dec 2018 18:57:42 +0000 (13:57 -0500)]
XArray: Fix xa_alloc when id exceeds max

Specifying a starting ID greater than the maximum ID isn't something
attempted very often, but it should fail.  It was succeeding due to
xas_find_marked() returning the wrong error state, so add tests for
both xa_alloc() and xas_find_marked().

Fixes: b803b42823d0 ("xarray: Add XArray iterators")
Signed-off-by: Matthew Wilcox <willy@infradead.org>
5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Thu, 13 Dec 2018 19:02:23 +0000 (11:02 -0800)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Doug Ledford:
 "We have 5 small fixes for this pull request. One is a performance
  regression, so not necessarily strictly a fix, but it was small and
  reasonable and claimed to avoid thrashing in the scheduler, so I took
  it. The remaining are all legitimate fixes that match the "we take
  fixes any time" criteria.

  Summary:

   - One performance regression for hfi1

   - One kasan fix for hfi1

   - A couple mlx5 fixes

   - A core oops fix"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  IB/core: Fix oops in netdev_next_upper_dev_rcu()
  IB/mlx5: Block DEVX umem from the non applicable cases
  IB/mlx5: Fix implicit ODP interrupted page fault
  IB/hfi1: Fix an out-of-bounds access in get_hw_stats
  IB/hfi1: Fix a latency issue for small messages

5 years agoMerge tag 'mmc-v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Thu, 13 Dec 2018 18:59:02 +0000 (10:59 -0800)]
Merge tag 'mmc-v4.20-rc5' of git://git./linux/kernel/git/ulfh/mmc

Pull mmc fixes from Ulf Hansson:
 "MMC core:
   - Fixup RPMB requests to use mrq->sbc when sending CMD23

  MMC host:
   - omap: Fix broken MMC/SD on OMAP15XX/OMAP5910/OMAP310

   - sdhci-omap: Fix DCRC error handling during tuning

   - sdhci: Fixup the timeout check window for clock and reset"

* tag 'mmc-v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: sdhci: fix the timeout check window for clock and reset
  mmc: sdhci-omap: Fix DCRC error handling during tuning
  MMC: OMAP: fix broken MMC on OMAP15XX/OMAP5910/OMAP310
  mmc: core: use mrq->sbc when sending CMD23 for RPMB

5 years agoMerge tag 'sound-4.20-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 13 Dec 2018 18:54:13 +0000 (10:54 -0800)]
Merge tag 'sound-4.20-rc7' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Only usual suspects here: a few more fixups for Realtek HD-audio on
  various PCs, including a regression fix in the previous fix for Lenovo
  X1 Carbon, as well as a typo fix in the recent Fireface patch"

* tag 'sound-4.20-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda/realtek: Enable audio jacks of ASUS UX433FN/UX333FA with ALC294
  ALSA: hda/realtek: Enable audio jacks of ASUS UX533FD with ALC294
  ALSA: hda/realtek: ALC294 mic and headset-mode fixups for ASUS X542UN
  ALSA: fireface: fix reference to wrong register for clock configuration
  ALSA: hda/realtek - Fix the mute LED regresion on Lenovo X1 Carbon
  ALSA: hda/realtek - Fixed headphone issue for ALC700

5 years agobcache: print number of keys in trace_bcache_journal_write
Guoju Fang [Thu, 13 Dec 2018 14:53:57 +0000 (22:53 +0800)]
bcache: print number of keys in trace_bcache_journal_write

Sometimes flush journal may be very frequent, so it's useful to dump
number of keys every time write journal.

Signed-off-by: Guoju Fang <fangguoju@gmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: set writeback_percent in a flexible range
Coly Li [Thu, 13 Dec 2018 14:53:56 +0000 (22:53 +0800)]
bcache: set writeback_percent in a flexible range

Because CUTOFF_WRITEBACK is defined as 40, so before the changes of
dynamic cutoff writeback values, writeback_percent is limited to [0,
CUTOFF_WRITEBACK]. Any value larger than CUTOFF_WRITEBACK will be fixed
up to 40.

Now cutof writeback limit is a dynamic value bch_cutoff_writeback, so
the range of writeback_percent can be a more flexible range as [0,
bch_cutoff_writeback]. The flexibility is, it can be expended to a
larger or smaller range than [0, 40], depends on how value
bch_cutoff_writeback is specified.

The default value is still strongly recommended to most of users for
most of workloads. But for people who want to do research on bcache
writeback perforamnce tuning, they may have chance to specify more
flexible writeback_percent in range [0, 70].

Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: make cutoff_writeback and cutoff_writeback_sync tunable
Coly Li [Thu, 13 Dec 2018 14:53:55 +0000 (22:53 +0800)]
bcache: make cutoff_writeback and cutoff_writeback_sync tunable

Currently the cutoff writeback and cutoff writeback sync thresholds are
defined by CUTOFF_WRITEBACK (40) and CUTOFF_WRITEBACK_SYNC (70) as
static values. Most of time these they work fine, but when people want
to do research on bcache writeback mode performance tuning, there is no
chance to modify the soft and hard cutoff writeback values.

This patch introduces two module parameters bch_cutoff_writeback_sync
and bch_cutoff_writeback which permit people to tune the values when
loading bcache.ko. If they are not specified by module loading, current
values CUTOFF_WRITEBACK_SYNC and CUTOFF_WRITEBACK will be used as
default and nothing changes.

When people want to tune this two values,
- cutoff_writeback can be set in range [1, 70]
- cutoff_writeback_sync can be set in range [1, 90]
- cutoff_writeback always <= cutoff_writeback_sync

The default values are strongly recommended to most of users for most of
workloads. Anyway, if people wants to take their own risk to do research
on new writeback cutoff tuning for their own workload, now they can make
it.

Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: add MODULE_DESCRIPTION information
Coly Li [Thu, 13 Dec 2018 14:53:54 +0000 (22:53 +0800)]
bcache: add MODULE_DESCRIPTION information

This patch moves MODULE_AUTHOR and MODULE_LICENSE to end of super.c, and
add MODULE_DESCRIPTION("Bcache: a Linux block layer cache").

This is preparation for adding module parameters.

Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: option to automatically run gc thread after writeback
Coly Li [Thu, 13 Dec 2018 14:53:53 +0000 (22:53 +0800)]
bcache: option to automatically run gc thread after writeback

The option gc_after_writeback is disabled by default, because garbage
collection will discard SSD data which drops cached data.

Echo 1 into /sys/fs/bcache/<UUID>/internal/gc_after_writeback will
enable this option, which wakes up gc thread when writeback accomplished
and all cached data is clean.

This option is helpful for people who cares writing performance more. In
heavy writing workload, all cached data can be clean only happens when
writeback thread cleans all cached data in I/O idle time. In such
situation a following gc running may help to shrink bcache B+ tree and
discard more clean data, which may be helpful for future writing
requests.

If you are not sure whether this is helpful for your own workload,
please leave it as disabled by default.

Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: introduce force_wake_up_gc()
Coly Li [Thu, 13 Dec 2018 14:53:52 +0000 (22:53 +0800)]
bcache: introduce force_wake_up_gc()

Garbage collection thread starts to work when c->sectors_to_gc is
negative value, otherwise nothing will happen even the gc thread is
woken up by wake_up_gc().

force_wake_up_gc() sets c->sectors_to_gc to -1 before calling
wake_up_gc(), then gc thread may have chance to run if no one else sets
c->sectors_to_gc to a positive value before gc_should_run().

This routine can be called where the gc thread is woken up and required
to run in force.

Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: cannot set writeback_running via sysfs if no writeback kthread created
Shenghui Wang [Thu, 13 Dec 2018 14:53:51 +0000 (22:53 +0800)]
bcache: cannot set writeback_running via sysfs if no writeback kthread created

"echo 1 > writeback_running" marks writeback_running even if no
writeback kthread created as "d_strtoul(writeback_running)" will simply
set dc-> writeback_running without checking the existence of
dc->writeback_thread.

Add check for setting writeback_running via sysfs: if no writeback
kthread available, reject setting to 1.

v2 -> v3:
  * Make message on wrong assignment more clear.
  * Print name of bcache device instead of name of backing device.

Signed-off-by: Shenghui Wang <shhuiw@foxmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: do not mark writeback_running too early
Shenghui Wang [Thu, 13 Dec 2018 14:53:50 +0000 (22:53 +0800)]
bcache: do not mark writeback_running too early

A fresh backing device is not attached to any cache_set, and
has no writeback kthread created until first attached to some
cache_set.

But bch_cached_dev_writeback_init run
"
dc->writeback_running = true;
WARN_ON(test_and_clear_bit(BCACHE_DEV_WB_RUNNING,
&dc->disk.flags));
"
for any newly formatted backing devices.

For a fresh standalone backing device, we can get something like
following even if no writeback kthread created:
------------------------
/sys/block/bcache0/bcache# cat writeback_running
1
/sys/block/bcache0/bcache# cat writeback_rate_debug
rate: 512.0k/sec
dirty: 0.0k
target: 0.0k
proportional: 0.0k
integral: 0.0k
change: 0.0k/sec
next io: -15427384ms

The none ZERO fields are misleading as no alive writeback kthread yet.

Set dc->writeback_running false as no writeback thread created in
bch_cached_dev_writeback_init().

We have writeback thread created and woken up in bch_cached_dev_writeback
_start(). Set dc->writeback_running true before bch_writeback_queue()
called, as a writeback thread will check if dc->writeback_running is true
before writing back dirty data, and hung if false detected.

After the change, we can get the following output for a fresh standalone
backing device:
-----------------------
/sys/block/bcache0/bcache$ cat writeback_running
0
/sys/block/bcache0/bcache# cat writeback_rate_debug
rate: 0.0k/sec
dirty: 0.0k
target: 0.0k
proportional: 0.0k
integral: 0.0k
change: 0.0k/sec
next io: 0ms

v1 -> v2:
  Set dc->writeback_running before bch_writeback_queue() called,

Signed-off-by: Shenghui Wang <shhuiw@foxmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: update comment in sysfs.c
Shenghui Wang [Thu, 13 Dec 2018 14:53:49 +0000 (22:53 +0800)]
bcache: update comment in sysfs.c

We have struct cached_dev allocated by kzalloc in register_bcache(),
which initializes all the fields of cached_dev with 0s. And commit
ce4c3e19e520 ("bcache: Replace bch_read_string_list() by
__sysfs_match_string()") has remove the string "default".

Update the comment.

Signed-off-by: Shenghui Wang <shhuiw@foxmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: update comment for bch_data_insert
Shenghui Wang [Thu, 13 Dec 2018 14:53:48 +0000 (22:53 +0800)]
bcache: update comment for bch_data_insert

commit 220bb38c21b8 ("bcache: Break up struct search") introduced
changes to struct search and s->iop. bypass/bio are fields of struct
data_insert_op now. Update the comment.

Signed-off-by: Shenghui Wang <shhuiw@foxmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: do not check if debug dentry is ERR or NULL explicitly on remove
Shenghui Wang [Thu, 13 Dec 2018 14:53:47 +0000 (22:53 +0800)]
bcache: do not check if debug dentry is ERR or NULL explicitly on remove

debugfs_remove and debugfs_remove_recursive will check if the dentry
pointer is NULL or ERR, and will do nothing in that case.

Remove the check in cache_set_free and bch_debug_init.

Signed-off-by: Shenghui Wang <shhuiw@foxmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agobcache: add comment for cache_set->fill_iter
Shenghui Wang [Thu, 13 Dec 2018 14:53:46 +0000 (22:53 +0800)]
bcache: add comment for cache_set->fill_iter

We have the following define for btree iterator:
struct btree_iter {
size_t size, used;
#ifdef CONFIG_BCACHE_DEBUG
struct btree_keys *b;
#endif
struct btree_iter_set {
struct bkey *k, *end;
} data[MAX_BSETS];
};

We can see that the length of data[] field is static MAX_BSETS, which is
defined as 4 currently.

But a btree node on disk could have too many bsets for an iterator to fit
on the stack - maybe far more that MAX_BSETS. Have to dynamically allocate
space to host more btree_iter_sets.

bch_cache_set_alloc() will make sure the pool cache_set->fill_iter can
allocate an iterator equipped with enough room that can host
(sb.bucket_size / sb.block_size)
btree_iter_sets, which is more than static MAX_BSETS.

bch_btree_node_read_done() will use that pool to allocate one iterator, to
host many bsets in one btree node.

Add more comment around cache_set->fill_iter to make code less confusing.

Signed-off-by: Shenghui Wang <shhuiw@foxmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agodrm/vmwgfx: Protect from excessive execbuf kernel memory allocations v3
Thomas Hellstrom [Wed, 12 Dec 2018 10:52:08 +0000 (11:52 +0100)]
drm/vmwgfx: Protect from excessive execbuf kernel memory allocations v3

With the new validation code, a malicious user-space app could
potentially submit command streams with enough buffer-object and resource
references in them to have the resulting allocated validion nodes and
relocations make the kernel run out of GFP_KERNEL memory.

Protect from this by having the validation code reserve TTM graphics
memory when allocating.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Deepak Rawat <drawat@vmware.com>
---
v2: Removed leftover debug printouts

5 years agonvme-rdma: support separate queue maps for read and write
Sagi Grimberg [Wed, 12 Dec 2018 07:38:58 +0000 (23:38 -0800)]
nvme-rdma: support separate queue maps for read and write

llow NVMF_OPT_NR_WRITE_QUEUES to describe additional write queues.  In
addition, implement .map_queues that will apply 2 queue maps for read
and write queue sets.

Note that with the separate queue map, HCTX_TYPE_READ will always use
nr_io_queues and HCTX_TYPE_DEFAULT will use nr_write_queues.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme-tcp: support separate queue maps for read and write
Sagi Grimberg [Wed, 12 Dec 2018 07:38:57 +0000 (23:38 -0800)]
nvme-tcp: support separate queue maps for read and write

Allow NVMF_OPT_NR_WRITE_QUEUES to describe additional write queues.  In
addition, implement .map_queues that will apply 2 queue maps for read
and write queue sets.

Note that with the separate queue map, HCTX_TYPE_READ will always use
nr_io_queues and HCTX_TYPE_DEFAULT will use nr_write_queues.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme-fabrics: allow user to set nr_write_queues for separate queue maps
Sagi Grimberg [Wed, 12 Dec 2018 07:38:56 +0000 (23:38 -0800)]
nvme-fabrics: allow user to set nr_write_queues for separate queue maps

This argument will specify how many I/O queues will be connected in
create_ctrl in addition to nr_io_queues. With this configuration, I/O
that carries payload from the host to the target, will use the default
hctx queue map, and I/O that involves target to host transfers will use
the read hctx queue map.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme-fabrics: add missing nvmf_ctrl_options documentation
Sagi Grimberg [Wed, 12 Dec 2018 07:38:55 +0000 (23:38 -0800)]
nvme-fabrics: add missing nvmf_ctrl_options documentation

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agoblk-mq-rdma: pass in queue map to blk_mq_rdma_map_queues
Sagi Grimberg [Wed, 12 Dec 2018 07:38:54 +0000 (23:38 -0800)]
blk-mq-rdma: pass in queue map to blk_mq_rdma_map_queues

Will be used by nvme-rdma for queue map separation support.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: update smart log with num err log entries
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:48 +0000 (15:11 -0800)]
nvmet: update smart log with num err log entries

Now that we have error log page implementation update smart log command
handler to provide number of error log entries in the lifetime of the
controller field.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error log page cmd handler
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:47 +0000 (15:11 -0800)]
nvmet: add error log page cmd handler

Now that we have support for all the major parts of the target we add
a NVMe error log page handler so that host can read the log page.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error log support for file backend
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:43 +0000 (15:11 -0800)]
nvmet: add error log support for file backend

This patch adds support for the file backend to populate the
error log entries. Here we map the errno to the NVMe status codes.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error log support for bdev backend
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:42 +0000 (15:11 -0800)]
nvmet: add error log support for bdev backend

This patch adds the support for the block device backend to populate the
error log entries. Here we map the blk_status_t to the NVMe status.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error log support for admin-cmd
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:46 +0000 (15:11 -0800)]
nvmet: add error log support for admin-cmd

This patch adds the support to maintain the error log page for admin
commands.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error log support for rdma backend
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:45 +0000 (15:11 -0800)]
nvmet: add error log support for rdma backend

This patch adds the support to maintain the error log page for rdma
transport, we mainly focus here on the NVME_INVALID_FIELD errors.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error log support for fabrics-cmd
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:44 +0000 (15:11 -0800)]
nvmet: add error log support for fabrics-cmd

This patch adds the support to maintain error log page for the fabrics
prop get, prop set, and admin connect commands. Here we also update the
discovery.c and add update set/get features and parse functions to
support error log page.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error log support in the core
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:41 +0000 (15:11 -0800)]
nvmet: add error log support in the core

This patch adds the support to maintain error log page for the
nvmet-core.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add interface to update error-log page
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:40 +0000 (15:11 -0800)]
nvmet: add interface to update error-log page

This patch adds nvmet_req based interface to the nvmet-core so that
we can update the error log page. We update error log page in
the request completion path when status is not set to NVME_SC_SUCCESS.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: add error-log definitions
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:39 +0000 (15:11 -0800)]
nvmet: add error-log definitions

This patch adds necessary fields in the target data structures to
support error log page. For a target controller, we add a new error log
field to maintain the error log, at any given point we maintain error
entries equal to NVMET_ERROR_LOG_SLOTS for each controller. In the
following patch, we also update the error log page entry in the I/O
completion path so we introduce a spinlock for synchronization of the
log.

For nvmet_req, we add a new field error_loc to hold the location of
the error in the command when the actual error occurs for each request
and a starting LBA if applicable.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme: add error log page slot definition
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:38 +0000 (15:11 -0800)]
nvme: add error log page slot definition

This patch adds the NVMe error slot definition from the spec.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme: remove nvme_common command cdw10 array
Chaitanya Kulkarni [Wed, 12 Dec 2018 23:11:37 +0000 (15:11 -0800)]
nvme: remove nvme_common command cdw10 array

This is a preparation patch which removes the nvme common command cdw10
array and replace with individual fields. This is needed for the nvmet
error log page implementation make is error log page entry offset
assignment easier.

Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: remove unused variable
Sagi Grimberg [Thu, 13 Dec 2018 07:01:54 +0000 (23:01 -0800)]
nvmet: remove unused variable

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme: provide fallback for discard alloc failure
Jens Axboe [Wed, 12 Dec 2018 16:18:11 +0000 (09:18 -0700)]
nvme: provide fallback for discard alloc failure

When boxes are run near (or to) OOM, we have a problem with the discard
page allocation in nvme. If we fail allocating the special page, we
return busy, and it'll get retried. But since ordering is honored for
dispatch requests, we can keep retrying this same IO and failing. Behind
that IO could be requests that want to free memory, but they never get
the chance.

Allocate a fixed discard page per controller for a safe fallback, and use
that if the initial allocation fails.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme: add __exit annotation
Chengguang Xu [Mon, 10 Dec 2018 23:24:34 +0000 (07:24 +0800)]
nvme: add __exit annotation

Add __exit annotation to cleanup helper which is only
called once in the module.

Signed-off-by: Chengguang Xu <cgxu519@gmx.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme-tcp: add NVMe over TCP host driver
Sagi Grimberg [Tue, 4 Dec 2018 01:52:17 +0000 (17:52 -0800)]
nvme-tcp: add NVMe over TCP host driver

This patch implements the NVMe over TCP host driver. It can be used to
connect to remote NVMe over Fabrics subsystems over good old TCP/IP.

The driver implements the TP 8000 of how nvme over fabrics capsules and
data are encapsulated in nvme-tcp pdus and exchaged on top of a TCP byte
stream. nvme-tcp header and data digest are supported as well.

To connect to all NVMe over Fabrics controllers reachable on a given taget
port over TCP use the following command:

nvme connect-all -t tcp -a $IPADDR

This requires the latest version of nvme-cli with TCP support.

Signed-off-by: Sagi Grimberg <sagi@lightbitslabs.com>
Signed-off-by: Roy Shterman <roys@lightbitslabs.com>
Signed-off-by: Solganik Alexander <sashas@lightbitslabs.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet: allow configfs tcp trtype configuration
Sagi Grimberg [Tue, 4 Dec 2018 01:52:16 +0000 (17:52 -0800)]
nvmet: allow configfs tcp trtype configuration

Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Sagi Grimberg <sagi@lightbitslabs.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvmet-tcp: add NVMe over TCP target driver
Sagi Grimberg [Tue, 4 Dec 2018 01:52:15 +0000 (17:52 -0800)]
nvmet-tcp: add NVMe over TCP target driver

This patch implements the TCP transport driver for the NVMe over Fabrics
target stack. This allows exporting NVMe over Fabrics functionality over
good old TCP/IP.

The driver implements the TP 8000 of how nvme over fabrics capsules and
data are encapsulated in nvme-tcp pdus and exchaged on top of a TCP byte
stream. nvme-tcp header and data digest are supported as well.

Signed-off-by: Sagi Grimberg <sagi@lightbitslabs.com>
Signed-off-by: Roy Shterman <roys@lightbitslabs.com>
Signed-off-by: Solganik Alexander <sashas@lightbitslabs.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme-tcp: Add protocol header
Sagi Grimberg [Tue, 4 Dec 2018 01:52:14 +0000 (17:52 -0800)]
nvme-tcp: Add protocol header

Signed-off-by: Sagi Grimberg <sagi@lightbitslabs.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme-fabrics: allow user passing data digest
Sagi Grimberg [Tue, 4 Dec 2018 01:52:13 +0000 (17:52 -0800)]
nvme-fabrics: allow user passing data digest

Data digest is a nvme-tcp specific feature, but nothing prevents other
transports reusing the concept so do not associate with tcp transport
solely.

Signed-off-by: Sagi Grimberg <sagi@lightbitslabs.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme-fabrics: allow user passing header digest
Sagi Grimberg [Tue, 4 Dec 2018 01:52:12 +0000 (17:52 -0800)]
nvme-fabrics: allow user passing header digest

Header digest is a nvme-tcp specific feature, but nothing prevents other
transports reusing the concept so do not associate with tcp transport
solely.

Signed-off-by: Sagi Grimberg <sagi@lightbitslabs.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>