'active_io' is used for mddev_suspend() and it's initialized in
md_run(), this restrict that 'reconfig_mutex' must be held and
"mddev->pers" must be set before calling mddev_suspend().

Initialize 'active_io' early so that mddev_suspend() is safe to call
once mddev is allocated, this will be helpful to refactor
mddev_suspend() in following patches.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825030956.1527023-2-yukuai1@huaweicloud.com

Before this patch, for read-only array:

md_check_recovery() check that 'MD_RECOVERY_NEEDED' is set, then it will
call remove_and_add_spares() directly to try to remove and add rdevs
from array.

After this patch:

1) md_check_recovery() check that 'MD_RECOVERY_NEEDED' is set, and the
   worker 'sync_work' is not pending, and there are rdevs can be added
   or removed, then it will queue new work md_start_sync();
2) md_start_sync() will call remove_and_add_spares() and exist;

This change make sure that array reconfiguration is independent from
daemon, and it'll be much easier to synchronize it with io, consier
that io may rely on daemon thread to be done.

Also fix a problem that 'pers->spars_active' is called after
remove_and_add_spares(), which order is wrong, because spares must
active first, and then remove_and_add_spares() can add spares to the
array, like what read-write case does:

1) daemon set 'MD_RECOVERY_RUNNING', register new sync thread to do
   recovery;
2) recovery is done, md_do_sync() set 'MD_RECOVERY_DONE' before return;
3) daemon call 'pers->spars_active', and clear 'MD_RECOVERY_RUNNING';
4) in the next round of daemon, call remove_and_add_spares() to add
   spares to the array.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825031622.1530464-8-yukuai1@huaweicloud.com

There are no functional changes, just to make the code simpler and
prepare to delay remove_and_add_spares() to md_start_sync().
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825031622.1530464-7-yukuai1@huaweicloud.com

There are no functional changes, just to make the code simpler and
prepare to delay remove_and_add_spares() to md_start_sync().
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825031622.1530464-6-yukuai1@huaweicloud.com

There are no functional changes, just to make the code simpler and
prepare to delay remove_and_add_spares() to md_start_sync().
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825031622.1530464-5-yukuai1@huaweicloud.com

Before this patch, for read-write array:

1) md_check_recover() found that something need to be done, and it'll
   try to grab 'reconfig_mutex'. The case that md_check_recover() need
   to do something:
   - array is not suspend;
   - super_block need to be updated;
   - 'MD_RECOVERY_NEEDED' or 'MD_RECOVERY_DONE' is set;
   - unusual case related to safemode;

2) if 'MD_RECOVERY_RUNNING' is not set, and 'MD_RECOVERY_NEEDED' is set,
   md_check_recover() will try to choose a sync action, and then queue a
   work md_start_sync().

3) md_start_sync() register sync_thread;

After this patch,

1) is the same;
2) if 'MD_RECOVERY_RUNNING' is not set, and 'MD_RECOVERY_NEEDED' is set,
   queue a work md_start_sync() directly;
3) md_start_sync() will try to choose a sync action, and then register
   sync_thread();

Because 'MD_RECOVERY_RUNNING' is cleared when sync_thread is done, 2)
and 3) and md_do_sync() is always ran in serial and they can never
concurrent, this change should not introduce any behavior change for now.

Also fix a problem that md_start_sync() can clear 'MD_RECOVERY_RUNNING'
without protection in error path, which might affect the logical in
md_check_recovery().

The advantage to change this is that array reconfiguration is
independent from daemon now, and it'll be much easier to synchronize it
with io, consider that io may rely on daemon thread to be done.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825031622.1530464-4-yukuai1@huaweicloud.com

There are no functional changes, on the one hand make the code cleaner,
on the other hand prevent following checkpatch error in the next patch to
delay choosing sync action to md_start_sync().

ERROR: do not use assignment in if condition
+       } else if ((spares = remove_and_add_spares(mddev, NULL))) {
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825031622.1530464-3-yukuai1@huaweicloud.com

It's a little weird to borrow 'del_work' for md_start_sync(), declare
a new work_struct 'sync_work' for md_start_sync().
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20230825031622.1530464-2-yukuai1@huaweicloud.com

Add batched mq_ops.queue_rqs() support in null_blk for testing. The
implementation is much easy since null_blk doesn't have commit_rqs().

We simply handle each request one by one, if errors are encountered,
leave them in the passed in list and return back.

There is about 3.6% improvement in IOPS of fio/t/io_uring on null_blk
with hw_queue_depth=256 on my test VM, from 1.09M to 1.13M.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20230913151616.3164338-6-chengming.zhou@linux.devSigned-off-by: Jens Axboe <axboe@kernel.dk>

Now we update driver tags request table in blk_mq_get_driver_tag(),
so the driver that support queue_rqs() have to update that inflight
table by itself.

Move it to blk_mq_start_request(), which is a better place where
we setup the deadline for request timeout check. And it's just
where the request becomes inflight.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20230913151616.3164338-5-chengming.zhou@linux.devSigned-off-by: Jens Axboe <axboe@kernel.dk>

Since active requests have been accounted when allocate driver tags,
we can remove this limit now.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20230913151616.3164338-4-chengming.zhou@linux.devSigned-off-by: Jens Axboe <axboe@kernel.dk>

Since the previous patch change to only account active requests when
we really allocate the driver tag, the RQF_MQ_INFLIGHT can be removed
and no double account problem.

1. none elevator: flush request will use the first pending request's
   driver tag, won't double account.

2. other elevator: flush request will be accounted when allocate driver
   tag when issue, and will be unaccounted when it put the driver tag.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20230913151616.3164338-3-chengming.zhou@linux.devSigned-off-by: Jens Axboe <axboe@kernel.dk>

There is a limit that batched queue_rqs() can't work on shared tags
queue, since the account of active requests can't be done there.

Now we account the active requests only in blk_mq_get_driver_tag(),
which is not the time we get driver tag actually (with none elevator).

To support batched queue_rqs() on shared tags queue, we move the
account of active requests to where we get the driver tag:

1. none elevator: blk_mq_get_tags() and blk_mq_get_tag()
2. other elevator: __blk_mq_alloc_driver_tag()

This is clearer and match with the unaccount side, which just happen
when we put the driver tag.

The other good point is that we don't need RQF_MQ_INFLIGHT trick
anymore, which used to avoid double account of flush request.
Now we only account when actually get the driver tag, so all is good.
We will remove RQF_MQ_INFLIGHT in the next patch.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20230913151616.3164338-2-chengming.zhou@linux.devSigned-off-by: Jens Axboe <axboe@kernel.dk>

Pull x86 fixes from Ingo Molnar:
 "Misc fixes:

   - Fix an UV boot crash

   - Skip spurious ENDBR generation on _THIS_IP_

   - Fix ENDBR use in putuser() asm methods

   - Fix corner case boot crashes on 5-level paging

   - and fix a false positive WARNING on LTO kernels"

* tag 'x86-urgent-2023-09-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/purgatory: Remove LTO flags
  x86/boot/compressed: Reserve more memory for page tables
  x86/ibt: Avoid duplicate ENDBR in __put_user_nocheck*()
  x86/ibt: Suppress spurious ENDBR
  x86/platform/uv: Use alternate source for socket to node data

Pull scheduler fixes from Ingo Molnar:
 "Fix a performance regression on large SMT systems, an Intel SMT4
  balancing bug, and a topology setup bug on (Intel) hybrid processors"

* tag 'sched-urgent-2023-09-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/sched: Restore the SD_ASYM_PACKING flag in the DIE domain
  sched/fair: Fix SMT4 group_smt_balance handling
  sched/fair: Optimize should_we_balance() for large SMT systems

Pull objtool fix from Ingo Molnar:
 "Fix a cold functions related false-positive objtool warning that
  triggers on Clang"

* tag 'objtool-urgent-2023-09-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  objtool: Fix _THIS_IP_ detection for cold functions

Pull WARN fix from Ingo Molnar:
 "Fix a missing preempt-enable in the WARN() slowpath"

* tag 'core-urgent-2023-09-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  panic: Reenable preemption in WARN slowpath

The choose_32_64() macros were added to deal with an odd inconsistency
between the 32-bit and 64-bit layout of 'struct stat' way back when in
commit a52dd971 ("vfs: de-crapify "cp_new_stat()" function").

Then a decade later Mikulas noticed that said inconsistency had been a
mistake in the early x86-64 port, and shouldn't have existed in the
first place.  So commit 932aba1e ("stat: fix inconsistency between
struct stat and struct compat_stat") removed the uses of the helpers.

But the helpers remained around, unused.

Get rid of them.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull smb client fixes from Steve French:
 "Three small SMB3 client fixes, one to improve a null check and two
  minor cleanups"

* tag '6.6-rc1-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  smb3: fix some minor typos and repeated words
  smb3: correct places where ENOTSUPP is used instead of preferred EOPNOTSUPP
  smb3: move server check earlier when setting channel sequence number

Pull smb server fixes from Steve French:
 "Two ksmbd server fixes"

* tag '6.6-rc1-ksmbd' of git://git.samba.org/ksmbd:
  ksmbd: fix passing freed memory 'aux_payload_buf'
  ksmbd: remove unneeded mark_inode_dirty in set_info_sec()

Pull ext4 fixes from Ted Ts'o:
 "Regression and bug fixes for ext4"

* tag 'ext4_for_linus-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix rec_len verify error
  ext4: do not let fstrim block system suspend
  ext4: move setting of trimmed bit into ext4_try_to_trim_range()
  jbd2: Fix memory leak in journal_init_common()
  jbd2: Remove page size assumptions
  buffer: Make bh_offset() work for compound pages

-flto* implies -ffunction-sections. With LTO enabled, ld.lld generates
multiple .text sections for purgatory.ro:

  $ readelf -S purgatory.ro  | grep " .text"
    [ 1] .text             PROGBITS         0000000000000000  00000040
    [ 7] .text.purgatory   PROGBITS         0000000000000000  000020e0
    [ 9] .text.warn        PROGBITS         0000000000000000  000021c0
    [13] .text.sha256_upda PROGBITS         0000000000000000  000022f0
    [15] .text.sha224_upda PROGBITS         0000000000000000  00002be0
    [17] .text.sha256_fina PROGBITS         0000000000000000  00002bf0
    [19] .text.sha224_fina PROGBITS         0000000000000000  00002cc0

This causes WARNING from kexec_purgatory_setup_sechdrs():

  WARNING: CPU: 26 PID: 110894 at kernel/kexec_file.c:919
  kexec_load_purgatory+0x37f/0x390

Fix this by disabling LTO for purgatory.

[ AFAICT, x86 is the only arch that supports LTO and purgatory. ]

We could also fix this with an explicit linker script to rejoin .text.*
sections back into .text. However, given the benefit of LTOing purgatory
is small, simply disable the production of more .text.* sections for now.

Fixes: b33fff07 ("x86, build: allow LTO to be selected")
Signed-off-by: Song Liu <song@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Sami Tolvanen <samitolvanen@google.com>
Link: https://lore.kernel.org/r/20230914170138.995606-1-song@kernel.org

The decompressor has a hard limit on the number of page tables it can
allocate. This limit is defined at compile-time and will cause boot
failure if it is reached.

The kernel is very strict and calculates the limit precisely for the
worst-case scenario based on the current configuration. However, it is
easy to forget to adjust the limit when a new use-case arises. The
worst-case scenario is rarely encountered during sanity checks.

In the case of enabling 5-level paging, a use-case was overlooked. The
limit needs to be increased by one to accommodate the additional level.
This oversight went unnoticed until Aaron attempted to run the kernel
via kexec with 5-level paging and unaccepted memory enabled.

Update wost-case calculations to include 5-level paging.

To address this issue, let's allocate some extra space for page tables.
128K should be sufficient for any use-case. The logic can be simplified
by using a single value for all kernel configurations.

[ Also add a warning, should this memory run low - by Dave Hansen. ]

Fixes: 34bbb000 ("x86/boot/compressed: Enable 5-level paging during decompression stage")
Reported-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20230915070221.10266-1-kirill.shutemov@linux.intel.com

Merge tag 'kbuild-fixes-v6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Fix kernel-devel RPM and linux-headers Deb package

 - Fix too long argument list error in 'make modules_install'

* tag 'kbuild-fixes-v6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kbuild: avoid long argument lists in make modules_install
  kbuild: fix kernel-devel RPM package and linux-headers Deb package

Commit 408579cd ("mm: Update do_vmi_align_munmap() return
semantics") seems to have updated one of the callers of do_vmi_munmap()
incorrectly: it used to check for the error case (which didn't
change: negative means error).

That commit changed the check to the success case (which did change:
before that commit, 0 was success, and 1 was "success and lock
downgraded".  After the change, it's always 0 for success, and the lock
will have been released if requested).

This didn't change any actual VM behavior _except_ for memory accounting
when 'VM_ACCOUNT' was set on the vma.  Which made the wrong return value
test fairly subtle, since everything continues to work.

Or rather - it continues to work but the "Committed memory" accounting
goes all wonky (Committed_AS value in /proc/meminfo), and depending on
settings that then causes problems much much later as the VM relies on
bogus statistics for its heuristics.

Revert that one line of the change back to the original logic.

Fixes: 408579cd ("mm: Update do_vmi_align_munmap() return semantics")
Reported-by: Christoph Biedl <linux-kernel.bfrz@manchmal.in-ulm.de>
Reported-bisected-and-tested-by: Michael Labiuk <michael.labiuk@virtuozzo.com>
Cc: Bagas Sanjaya <bagasdotme@gmail.com>
Cc: Liam R. Howlett <Liam.Howlett@oracle.com>
Link: https://lore.kernel.org/all/1694366957@msgid.manchmal.in-ulm.de/Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull SCSI fixes from James Bottomley:
 "16 small(ish) fixes all in drivers.

  The major fixes are in pm8001 (fixes MSI-X issue going back to its
  origin), the qla2xxx endianness fix, which fixes a bug on big endian
  and the lpfc ones which can cause an oops on module removal without
  them"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: lpfc: Prevent use-after-free during rmmod with mapped NVMe rports
  scsi: lpfc: Early return after marking final NLP_DROPPED flag in dev_loss_tmo
  scsi: lpfc: Fix the NULL vs IS_ERR() bug for debugfs_create_file()
  scsi: target: core: Fix target_cmd_counter leak
  scsi: pm8001: Setup IRQs on resume
  scsi: pm80xx: Avoid leaking tags when processing OPC_INB_SET_CONTROLLER_CONFIG command
  scsi: pm80xx: Use phy-specific SAS address when sending PHY_START command
  scsi: ufs: core: Poll HCS.UCRDY before issuing a UIC command
  scsi: ufs: core: Move __ufshcd_send_uic_cmd() outside host_lock
  scsi: qedf: Add synchronization between I/O completions and abort
  scsi: target: Replace strlcpy() with strscpy()
  scsi: qla2xxx: Fix NULL vs IS_ERR() bug for debugfs_create_dir()
  scsi: qla2xxx: Use raw_smp_processor_id() instead of smp_processor_id()
  scsi: qla2xxx: Correct endianness for rqstlen and rsplen
  scsi: ppa: Fix accidentally reversed conditions for 16-bit and 32-bit EPP
  scsi: megaraid_sas: Fix deadlock on firmware crashdump

Pull ata fixes from Damien Le Moal:

 - Fix link power management transitions to disallow unsupported states
   (Niklas)

 - A small string handling fix for the sata_mv driver (Christophe)

 - Clear port pending interrupts before reset, as per AHCI
   specifications (Szuying).

   Followup fixes for this one are to not clear ATA_PFLAG_EH_PENDING in
   ata_eh_reset() to allow EH to continue on with other actions recorded
   with error interrupts triggered before EH completes. And an
   additional fix to avoid thawing a port twice in EH (Niklas)

 - Small code style fixes in the pata_parport driver to silence the
   build bot as it keeps complaining about bad indentation (me)

 - A fix for the recent CDL code to avoid fetching sense data for
   successful commands when not necessary for correct operation (Niklas)

* tag 'ata-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal/libata:
  ata: libata-core: fetch sense data for successful commands iff CDL enabled
  ata: libata-eh: do not thaw the port twice in ata_eh_reset()
  ata: libata-eh: do not clear ATA_PFLAG_EH_PENDING in ata_eh_reset()
  ata: pata_parport: Fix code style issues
  ata: libahci: clear pending interrupt status
  ata: sata_mv: Fix incorrect string length computation in mv_dump_mem()
  ata: libata: disallow dev-initiated LPM transitions to unsupported states

Pull USB fix from Greg KH:
 "Here is a single USB fix for a much-reported regression for 6.6-rc1.

  It resolves a crash in the typec debugfs code for many systems. It's
  been in linux-next with no reported issues, and many people have
  reported it resolving their problem with 6.6-rc1"

* tag 'usb-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: typec: ucsi: Fix NULL pointer dereference

Pull driver core fixes from Greg KH:
 "Here is a single driver core fix for a much-reported-by-sysbot issue
  that showed up in 6.6-rc1. It's been submitted by many people, all in
  the same way, so it obviously fixes things for them all.

  Also in here is a single documentation update adding riscv to the
  embargoed hardware document in case there are any future issues with
  that processor family.

  Both of these have been in linux-next with no reported problems"

* tag 'driver-core-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  Documentation: embargoed-hardware-issues.rst: Add myself for RISC-V
  driver core: return an error when dev_set_name() hasn't happened

Pull char/misc fix from Greg KH:
 "Here is a single patch for 6.6-rc2 that reverts a 6.5 change for the
  comedi subsystem that has ended up being incorrect and caused drivers
  that were working for people to be unable to be able to be selected to
  build at all.

  To fix this, the Kconfig change needs to be reverted and a future set
  of fixes for the ioport dependancies will show up in 6.7-rc1 (there's
  no rush for them.)

  This has been in linux-next with no reported issues"

* tag 'char-misc-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  Revert "comedi: add HAS_IOPORT dependencies"

Pull i2c fixes from Wolfram Sang:
 "The main thing is the removal of 'probe_new' because all i2c client
  drivers are converted now. Thanks Uwe, this marks the end of a long
  conversion process.

  Other than that, we have a few Kconfig updates and driver bugfixes"

* tag 'i2c-for-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: cadence: Fix the kernel-doc warnings
  i2c: aspeed: Reset the i2c controller when timeout occurs
  i2c: I2C_MLXCPLD on ARM64 should depend on ACPI
  i2c: Make I2C_ATR invisible
  i2c: Drop legacy callback .probe_new()
  w1: ds2482: Switch back to use struct i2c_driver's .probe()

Currently, we fetch sense data for a _successful_ command if either:
1) Command was NCQ and ATA_DFLAG_CDL_ENABLED flag set (flag
   ATA_DFLAG_CDL_ENABLED will only be set if the Successful NCQ command
   sense data supported bit is set); or
2) Command was non-NCQ and regular sense data reporting is enabled.

This means that case 2) will trigger for a non-NCQ command which has
ATA_SENSE bit set, regardless if CDL is enabled or not.

This decision was by design. If the device reports that it has sense data
available, it makes sense to fetch that sense data, since the sk/asc/ascq
could be important information regardless if CDL is enabled or not.

However, the fetching of sense data for a successful command is done via
ATA EH. Considering how intricate the ATA EH is, we really do not want to
invoke ATA EH unless absolutely needed.

Before commit 18bd7718 ("scsi: ata: libata: Handle completion of CDL
commands using policy 0xD") we never fetched sense data for successful
commands.

In order to not invoke the ATA EH unless absolutely necessary, even if the
device claims support for sense data reporting, only fetch sense data for
successful (NCQ and non-NCQ commands) commands that are using CDL.

[Damien] Modified the check to test the qc flag ATA_QCFLAG_HAS_CDL
instead of the device support for CDL, which is implied for commands
using CDL.

Fixes: 3ac873c7 ("ata: libata-core: fix when to fetch sense data for successful commands")
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>

commit 1e641060 ("libata: clear eh_info on reset completion") added
a workaround that broke the retry mechanism in ATA EH.

Tejun himself suggested to remove this workaround when it was identified
to cause additional problems:
https://lore.kernel.org/linux-ide/20110426135027.GI878@htj.dyndns.org/

He even said:
"Hmm... it seems I wasn't thinking straight when I added that work around."
https://lore.kernel.org/linux-ide/20110426155229.GM878@htj.dyndns.org/

While removing the workaround solved the issue, however, the workaround was
kept to avoid "spurious hotplug events during reset", and instead another
workaround was added on top of the existing workaround in commit
8c56cacc ("libata: fix unexpectedly frozen port after ata_eh_reset()").

Because these IRQs happened when the port was frozen, we know that they
were actually a side effect of PxIS and IS.IPS(x) not being cleared before
the COMRESET. This is now done in commit 94152042eaa9 ("ata: libahci: clear
pending interrupt status"), so these workarounds can now be removed.

Since commit 1e641060 ("libata: clear eh_info on reset completion") has
now been reverted, the ATA EH retry mechanism is functional again, so there
is once again no need to thaw the port more than once in ata_eh_reset().

This reverts "the workaround on top of the workaround" introduced in commit
8c56cacc ("libata: fix unexpectedly frozen port after ata_eh_reset()").
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>

ata_scsi_port_error_handler() starts off by clearing ATA_PFLAG_EH_PENDING,
before calling ap->ops->error_handler() (without holding the ap->lock).

If an error IRQ is received while ap->ops->error_handler() is running,
the irq handler will set ATA_PFLAG_EH_PENDING.

Once ap->ops->error_handler() returns, ata_scsi_port_error_handler()
checks if ATA_PFLAG_EH_PENDING is set, and if it is, another iteration
of ATA EH is performed.

The problem is that ATA_PFLAG_EH_PENDING is not only cleared by
ata_scsi_port_error_handler(), it is also cleared by ata_eh_reset().

ata_eh_reset() is called by ap->ops->error_handler(). This additional
clearing done by ata_eh_reset() breaks the whole retry logic in
ata_scsi_port_error_handler(). Thus, if an error IRQ is received while
ap->ops->error_handler() is running, the port will currently remain
frozen and will never get re-enabled.

The additional clearing in ata_eh_reset() was introduced in commit
1e641060 ("libata: clear eh_info on reset completion").

Looking at the original error report:
https://marc.info/?l=linux-ide&m=124765325828495&w=2

We can see the following happening:
[    1.074659] ata3: XXX port freeze
[    1.074700] ata3: XXX hardresetting link, stopping engine
[    1.074746] ata3: XXX flipping SControl

[    1.411471] ata3: XXX irq_stat=400040 CONN|PHY
[    1.411475] ata3: XXX port freeze

[    1.420049] ata3: XXX starting engine
[    1.420096] ata3: XXX rc=0, class=1
[    1.420142] ata3: XXX clearing IRQs for thawing
[    1.420188] ata3: XXX port thawed
[    1.420234] ata3: SATA link up 3.0 Gbps (SStatus 123 SControl 300)

We are not supposed to be able to receive an error IRQ while the port is
frozen (PxIE is set to 0, i.e. all IRQs for the port are disabled).

AHCI 1.3.1 section 10.7.1.1 First Tier (IS Register) states:
"Each bit location can be thought of as reporting a '1' if the virtual
"interrupt line" for that port is indicating it wishes to generate an
interrupt. That is, if a port has one or more interrupt status bit set,
and the enables for those status bits are set, then this bit shall be set."

Additionally, AHCI state P:ComInit clearly shows that the state machine
will only jump to P:ComInitSetIS (which sets IS.IPS(x) to '1'), if PxIE.PCE
is set to '1'. In our case, PxIE is set to 0, so IS.IPS(x) won't get set.

So IS.IPS(x) only gets set if PxIS and PxIE is set.

AHCI 1.3.1 section 10.7.1.1 First Tier (IS Register) also states:
"The bits in this register are read/write clear. It is set by the level of
the virtual interrupt line being a set, and cleared by a write of '1' from
the software."

So if IS.IPS(x) is set, you need to explicitly clear it by writing a 1 to
IS.IPS(x) for that port.

Since PxIE is cleared, the only way to get an interrupt while the port is
frozen, is if IS.IPS(x) is set, and the only way IS.IPS(x) can be set when
the port is frozen, is if it was set before the port was frozen.

However, since commit 737dd811 ("ata: libahci: clear pending interrupt
status"), we clear both PxIS and IS.IPS(x) after freezing the port, but
before the COMRESET, so the problem that commit 1e641060 ("libata:
clear eh_info on reset completion") fixed can no longer happen.

Thus, revert commit 1e641060 ("libata: clear eh_info on reset
completion"), so that the retry logic in ata_scsi_port_error_handler()
works once again. (The retry logic is still needed, since we can still
get an error IRQ _after_ the port has been thawed, but before
ata_scsi_port_error_handler() takes the ap->lock in order to check
if ATA_PFLAG_EH_PENDING is set.)
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>

Merge tag 'linux-kselftest-fixes-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest

Pull more kselftest fixes from Shuah Khan
 "Fixes to user_events test and ftrace test.

  The user_events test was enabled by default in Linux 6.6-rc1. The
  following fixes are for bugs found since then:

   - add checks for dependencies and skip the test if they aren't met.

     The user_events test requires root access, and tracefs and
     user_events enabled. It leaves tracefs mounted and a fix is in
     progress for that missing piece.

   - create user_events test-specific Kconfig fragments

  ftrace test fixes:

   - unmount tracefs for recovering environment. Fix identified during
     the above mentioned user_events dependencies fix.

   - adds softlink to latest log directory improving usage"

* tag 'linux-kselftest-fixes-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  selftests: tracing: Fix to unmount tracefs for recovering environment
  selftests: user_events: create test-specific Kconfig fragments
  ftrace/selftests: Add softlink to latest log directory
  selftests/user_events: Fix failures when user_events is not installed

Pull nfsd fixes from Chuck Lever:

 - Use correct order when encoding NFSv4 RENAME change_info

 - Fix a potential oops during NFSD shutdown

* tag 'nfsd-6.6-1' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux:
  NFSD: fix possible oops when nfsd/pool_stats is closed.
  nfsd: fix change_info in NFSv4 RENAME replies

Pull power management fixes from Rafael Wysocki:
 "Fix the handling of block devices in the test_resume mode of
  hibernation (Chen Yu)"

* tag 'pm-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM: hibernate: Fix the exclusive get block device in test_resume mode
  PM: hibernate: Rename function parameter from snapshot_test to exclusive

Pull thermal control fixes from Rafael Wysocki:
 "These fix a thermal core breakage introduced by one of the recent
  changes, amend those changes by adding 'const' to a new callback
  argument and fix two memory leaks.

  Specifics:

   - Unbreak disabled trip point check in handle_thermal_trip() that may
     cause it to skip enabled trip points (Rafael Wysocki)

   - Add missing of_node_put() to of_find_trip_id() and
     thermal_of_for_each_cooling_maps() that each break out of a
     for_each_child_of_node() loop without dropping the reference to the
     child object (Julia Lawall)

   - Constify the recently added trip argument of the .get_trend()
     thermal zone callback (Rafael Wysocki)"

* tag 'thermal-6.6-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  thermal: core: Fix disabled trip point check in handle_thermal_trip()
  thermal: Constify the trip argument of the .get_trend() zone callback
  thermal/of: add missing of_node_put()

Merge tag 'for-6.6/dm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - Fix DM core retrieve_deps() UAF race due to missing locking of a DM
   table's list of devices that is managed using dm_{get,put}_device.

 - Revert DM core's half-baked RCU optimization if IO submitter has set
   REQ_NOWAIT. Can be revisited, and properly justified, after
   comprehensively auditing all of DM to also pass GFP_NOWAIT for any
   allocations if REQ_NOWAIT used.

* tag 'for-6.6/dm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm: don't attempt to queue IO under RCU protection
  dm: fix a race condition in retrieve_deps