Use a separate buffer for writing zeroes to the on-disk snapshot
exception store, make the updating of ps->current_area explicit and
refactor the code in preparation for the fix in the next patch.
No functional change.
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@kernel.org
The last_percent field is unused - remove it.
(It dates from when events were triggered as each X% filled up.)
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Fix a race condition with primary_pe ref_count handling.
put_pending_exception runs under dm_snapshot->lock, it does atomic_dec_and_test
on primary_pe->ref_count, and later does atomic_read primary_pe->ref_count.
__origin_write does atomic_dec_and_test on primary_pe->ref_count without holding
dm_snapshot->lock.
This opens the following race condition:
Assume two CPUs, CPU1 is executing put_pending_exception (and holding
dm_snapshot->lock). CPU2 is executing __origin_write in parallel.
primary_pe->ref_count == 2.
CPU1:
if (primary_pe && atomic_dec_and_test(&primary_pe->ref_count))
origin_bios = bio_list_get(&primary_pe->origin_bios);
... decrements primary_pe->ref_count to 1. Doesn't load origin_bios
CPU2:
if (first && atomic_dec_and_test(&primary_pe->ref_count)) {
flush_bios(bio_list_get(&primary_pe->origin_bios));
free_pending_exception(primary_pe);
/* If we got here, pe_queue is necessarily empty. */
return r;
}
... decrements primary_pe->ref_count to 0, submits pending bios, frees
primary_pe.
CPU1:
if (!primary_pe || primary_pe != pe)
free_pending_exception(pe);
... this has no effect.
if (primary_pe && !atomic_read(&primary_pe->ref_count))
free_pending_exception(primary_pe);
... sees ref_count == 0 (written by CPU 2), does double free !!
This bug can happen only if someone is simultaneously writing to both the
origin and the snapshot.
If someone is writing only to the origin, __origin_write will submit kcopyd
request after it decrements primary_pe->ref_count (so it can't happen that the
finished copy races with primary_pe->ref_count decrementation).
If someone is writing only to the snapshot, __origin_write isn't invoked at all
and the race can't happen.
The race happens when someone writes to the snapshot --- this creates
pending_exception with primary_pe == NULL and starts copying. Then, someone
writes to the same chunk in the snapshot, and __origin_write races with
termination of already submitted request in pending_complete (that calls
put_pending_exception).
This race may be reason for bugs:
http://bugzilla.kernel.org/show_bug.cgi?id=11636https://bugzilla.redhat.com/show_bug.cgi?id=465825
The patch fixes the code to make sure that:
1. If atomic_dec_and_test(&primary_pe->ref_count) returns false, the process
must no longer dereference primary_pe (because someone else may free it under
us).
2. If atomic_dec_and_test(&primary_pe->ref_count) returns true, the process
is responsible for freeing primary_pe.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Cc: stable@kernel.org
Write throughput to LVM snapshot origin volume is an order
of magnitude slower than those to LV without snapshots or
snapshot target volumes, especially in the case of sequential
writes with O_SYNC on.
The following patch originally written by Kevin Jamieson and
Jan Blunck and slightly modified for the current RCs by myself
tries to improve the performance by modifying the behaviour
of kcopyd, so that it pushes back an I/O job to the head of
the job queue instead of the tail as process_jobs() currently
does when it has to wait for free pages. This way, write
requests aren't shuffled to cause extra seeks.
I tested the patch against 2.6.27-rc5 and got the following results.
The test is a dd command writing to snapshot origin followed by fsync
to the file just created/updated. A couple of filesystem benchmarks
gave me similar results in case of sequential writes, while random
writes didn't suffer much.
dd if=/dev/zero of=<somewhere on snapshot origin> bs=4096 count=...
[conv=notrunc when updating]
1) linux 2.6.27-rc5 without the patch, write to snapshot origin,
average throughput (MB/s)
10M 100M 1000M
create,dd 511.46 610.72 11.81
create,dd+fsync 7.10 6.77 8.13
update,dd 431.63 917.41 12.75
update,dd+fsync 7.79 7.43 8.12
compared with write throughput to LV without any snapshots,
all dd+fsync and 1000 MiB writes perform very poorly.
10M 100M 1000M
create,dd 555.03 608.98 123.29
create,dd+fsync 114.27 72.78 76.65
update,dd 152.34 1267.27 124.04
update,dd+fsync 130.56 77.81 77.84
2) linux 2.6.27-rc5 with the patch, write to snapshot origin,
average throughput (MB/s)
10M 100M 1000M
create,dd 537.06 589.44 46.21
create,dd+fsync 31.63 29.19 29.23
update,dd 487.59 897.65 37.76
update,dd+fsync 34.12 30.07 26.85
Although still not on par with plain LV performance -
cannot be avoided because it's copy on write anyway -
this simple patch successfully improves throughtput
of dd+fsync while not affecting the rest.
Signed-off-by: Jan Blunck <jblunck@suse.de>
Signed-off-by: Kazuo Ito <ito.kazuo@oss.ntt.co.jp>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Cc: stable@kernel.org
To keep the size of changesets sane we split the switch by drivers;
to keep the damn thing bisectable we do the following:
1) rename the affected methods, add ones with correct
prototypes, make (few) callers handle both. That's this changeset.
2) for each driver convert to new methods. *ALL* drivers
are converted in this series.
3) kill the old (renamed) methods.
Note that it _is_ a flagday; all in-tree drivers are converted and by the
end of this series no trace of old methods remain. The only reason why
we do that this way is to keep the damn thing bisectable and allow per-driver
debugging if anything goes wrong.
New methods:
open(bdev, mode)
release(disk, mode)
ioctl(bdev, mode, cmd, arg) /* Called without BKL */
compat_ioctl(bdev, mode, cmd, arg)
locked_ioctl(bdev, mode, cmd, arg) /* Called with BKL, legacy */
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Analog of blkdev_driver_ioctl() with sane arguments. For
now uses fake struct file, by the end of the series it won't
and blkdev_driver_ioctl() will become a wrapper around it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The new extended partition support provides a much nicer was
to have partitions on md devices that the 'mdp' alternate major.
We cannot really get rid of 'mdp' at this time, but we can
enable extended partitions as that will probably make life
easier for sysadmins.
Signed-off-by: NeilBrown <neilb@suse.de>
The 'state' file for a device reports, for example, when the device
has failed. Changes should be reported to userspace ASAP without
the possibility of blocking on low-memory. sysfs_notify does
have that possibility (as it takes a mutex which can be held
across a kmalloc) so use sysfs_notify_dirent instead.
Signed-off-by: NeilBrown <neilb@suse.de>
Now that we have sysfs_notify_dirent, use it to notify changes
to md/array_state.
As sysfs_notify_dirent can be called in atomic context, we can
remove the delayed notify and the MD_NOTIFY_ARRAY_STATE flag.
Signed-off-by: NeilBrown <neilb@suse.de>
* 'for-linus' of git://neil.brown.name/md:
md: fix input truncation in safe_delay_store()
md: check for memory allocation failure in faulty personality
md: build failure due to missing delay.h
md: Relax minimum size restrictions on chunk_size.
md: remove space after function name in declaration and call.
md: Remove unnecessary #includes, #defines, and function declarations.
md: Convert remaining 1k representations in linear.c to sectors.
md: linear.c: Make two local variables sector-based.
md: linear: Represent dev_info->size and dev_info->offset in sectors.
md: linear.c: Remove broken debug code.
md: linear.c: Remove pointless initialization of curr_offset.
md: linear.c: Fix typo in comment.
md: Don't try to set an array to 'read-auto' if it is already in that state.
md: Allow metadata_version to be updated for externally managed metadata.
md: Fix rdev_size_store with size == 0
safe_delay_store() currently truncates the last character of input since
it tells strlcpy that the buffer can only hold 'len' characters, off by
one. sysfs already null terminates the buffer, so just increase the
last argument to strlcpy.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
It's a fault injection module, but I don't think we should oops here.
Signed-off-by: Sven Wegener <sven.wegener@stealer.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Neil Brown <neilb@suse.de>
Today's linux-next build (powerpc ppc64_defconfig) failed like this:
drivers/md/raid1.c: In function 'sync_request':
drivers/md/raid1.c:1759: error: implicit declaration of function 'msleep_interruptible'
make[3]: *** [drivers/md/raid1.o] Error 1
make[3]: *** Waiting for unfinished jobs....
drivers/md/raid10.c: In function 'sync_request':
drivers/md/raid10.c:1749: error: implicit declaration of function 'msleep_interruptible'
make[3]: *** [drivers/md/raid10.o] Error 1
drivers/md/md.c: In function 'md_do_sync':
drivers/md/md.c:5915: error: implicit declaration of function 'msleep'
Caused by commit 6caa3b0bbdb474647f6bdd8a958ffc46f78d8d58 ("md: Remove
unnecessary #includes, #defines, and function declarations"). I added
the following patch.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: NeilBrown <neilb@suse.de>
Multipath is best at handling transport errors. If it gets a device
error then there is not much the multipath layer can do. It will just
access the same device but from a different path.
This patch breaks up failfast into device, transport and driver errors.
The multipath layers (md and dm mutlipath) only ask the lower levels to
fast fail transport errors. The user of failfast, read ahead, will ask
to fast fail on all errors.
Note that blk_noretry_request will return true if any failfast bit
is set. This allows drivers that do not support the multipath failfast
bits to continue to fail on any failfast error like before. Drivers
like scsi that are able to fail fast specific errors can check
for the specific fail fast type. In the next patch I will convert
scsi.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Currently, the 'chunk_size' of an array must be at-least PAGE_SIZE.
This makes moving an array to a machine with a larger PAGE_SIZE, or
changing the kernel to use a larger PAGE_SIZE, can stop an array from
working.
For RAID10 and RAID4/5/6, this is non-trivial to fix as the resync
process works on whole pages at a time, and assumes them to be wholly
within a stripe. For other raid personalities, this restriction is
not needed at all and can be dropped.
So remove the test on chunk_size from common can, and add it in just
the places where it is needed: raid10 and raid4/5/6.
Signed-off-by: NeilBrown <neilb@suse.de>
Having
function (args)
instead of
function(args)
make is harder to search for calls of particular functions.
So remove all those spaces.
Signed-off-by: NeilBrown <neilb@suse.de>
This patch renames hash_spacing and preshift to spacing and
sector_shift respectively with the following change of semantics:
Case 1: (sizeof(sector_t) <= sizeof(u32)).
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In this case, we have sector_shift = preshift = 0 and spacing =
2 * hash_spacing.
Hence, the index for the hash table which is computed by the new code
in which_dev() as sector / spacing equals the old value which was
(sector/2) / hash_spacing.
Note also that the value of nb_zone stays the same because both sz
and base double.
Case 2: (sizeof(sector_t) > sizeof(u32)).
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(aka the shifting dance case). Here we have sector_shift = preshift +
1 and
spacing = 2 * hash_spacing
during the computation of nb_zone and curr_sector, but
spacing = hash_spacing
in which_dev() because in the last hunk of the patch for linear.c we
shift down conf->spacing (= 2 * hash_spacing) by one more bit than
in the old code.
Hence in the computation of nb_zone, sz and base have the same value
as before, so nb_zone is not affected. Also curr_sector in the next
hunk stays the same.
In which_dev() the hash table index is computed as
(sector >> sector_shift) / spacing
In view of sector_shift = preshift + 1 and spacing = hash_spacing,
this equals
((sector/2) >> preshift) / hash_spacing
which is the value computed by the old code.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: NeilBrown <neilb@suse.de>
This is a preparation for representing also the remaining fields of struct
linear_private_data as sectors.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: NeilBrown <neilb@suse.de>
Rename them to num_sectors and start_sector which is more descriptive.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: NeilBrown <neilb@suse.de>
conf->smallest_size is undefined since day one of the git repo..
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: NeilBrown <neilb@suse.de>
'read-auto' is a variant of 'readonly' which will switch to writable
on the first write attempt.
Calling do_md_stop to set the array readonly when it is already readonly
returns an error. So make sure not to do that.
Signed-off-by: NeilBrown <neilb@suse.de>
For externally managed metadata, the 'metadata_version' sysfs
attribute is really just a channel for user-space programs to
communicate about how the array is being managed.
It can be useful for this to be changed while the array is active.
Normally changes to metadata_version are not permitted while the array
is active. Change that so that if the metadata is externally managed,
the metadata_version can be changed to a different flavour of external
management.
Signed-off-by: NeilBrown <neilb@suse.de>
Fix rdev_size_store with size == 0.
size == 0 means to use the largest size allowed by the
underlying device and is used when modifying an active array.
This fixes a regression introduced by
commit d7027458d6
Cc: <stable@kernel.org>
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: NeilBrown <neilb@suse.de>
RAID autodetect has the side effect of requiring synchronisation
of all device drivers, which can make the boot several seconds longer
(I've measured 7 on one of my laptops).... even for systems that don't
have RAID setup for the root filesystem (the only FS where this matters).
This patch makes the default for autodetect a config option; either way
the user can always override via the kernel command line.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: NeilBrown <neilb@suse.de>
Detect and report buggy drivers that destroy their request_queue.
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Cc: Stefan Raspl <raspl@linux.vnet.ibm.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Publish dm_vcalloc in include/linux/device-mapper.h because this function is
used by targets.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Publish dm_table_unplug_all in include/linux/device-mapper.h because this
function is used by targets.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Publish dm_get_mapinfo in include/linux/device-mapper.h because this function
is used by targets.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Split struct dm_dev in two and publish the part that other targets need in
include/linux/device-mapper.h.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Don't wait between submitting crypt requests for a bio unless
we are short of memory.
There are two situations when we must split an encrypted bio:
1) there are no free pages;
2) the new bio would violate underlying device restrictions
(e.g. max hw segments).
In case (2) we do not need to wait.
Add output variable to crypt_alloc_buffer() to distinguish between
these cases.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Move the initialisation of ctx->pending into one place, at the
start of crypt_convert().
Introduce crypt_finished to indicate whether or not the encryption
is finished, for use in a later patch.
No functional change.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
The pending reference count must be incremented *before* the async work is
queued to another thread, not after. Otherwise there's a race if the
work completes and decrements the reference count before it gets incremented.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Make kcryptd_crypt_write_io_submit() responsible for decrementing
the pending count after an error.
Also fixes a bug in the async path that forgot to decrement it.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Make the caller reponsible for incrementing the pending count before calling
kcryptd_crypt_write_io_submit() in the non-async case to bring it into line
with the async case.
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Move kcryptd_crypt_write_convert_loop inside kcryptd_crypt_write_convert.
This change is needed for a later patch.
No functional change.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Factor out crypt io allocation code.
Later patches will call it from another place.
No functional change.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Move io pending to one place.
No functional change, usefull to simplify debugging.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Change uint32_t into chunk_t to remove 32-bit limitation on the
number of chunks on systems with 64-bit sector numbers.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Move this logic to a function, because it will be reused later.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
dm-raid1 is setting the 'DM_KCOPYD_IGNORE_ERROR' flag unconditionally
when assigning kcopyd work. kcopyd is responsible for copying an
assigned section of disk to one or more other disks. The
'DM_KCOPYD_IGNORE_ERROR' flag affects kcopyd in the following way:
When not set:
kcopyd will immediately stop the copy operation when an error is
encountered.
When set:
kcopyd will try to proceed regardless of errors and try to continue
copying any remaining amount.
Since dm-raid1 tracks regions of the address space that are (or
are not) in sync and it now has the ability to handle these
errors, we can safely enable this optimization. This optimization
is conditional on whether mirror error handling has been enabled.
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
This patch moves 'is_active' from struct dm_path to struct pgpath
as it does not need exporting.
Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
This patch allows path errors from the multipath ctr function to
propagate up to userspace as errno values from the ioctl() call.
This is in response to
https://www.redhat.com/archives/dm-devel/2008-May/msg00000.html
and
https://bugzilla.redhat.com/show_bug.cgi?id=444421
The patch only lets through the errors that it needs to in order to
get the path errors from parse_path().
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Since all bio_split calls refer the same single bio_split_pool, the bio_split
function can use bio_split_pool directly instead of the mempool_t parameter;
then the mempool_t parameter can be removed from bio_split param list, and
bio_split_pool is only referred in fs/bio.c file, can be marked static.
Signed-off-by: Denis ChengRq <crquan@gmail.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Move stats related fields - stamp, in_flight, dkstats - from disk to
part0 and unify stat handling such that...
* part_stat_*() now updates part0 together if the specified partition
is not part0. ie. part_stat_*() are now essentially all_stat_*().
* {disk|all}_stat_*() are gone.
* part_round_stats() is updated similary. It handles part0 stats
automatically and disk_round_stats() is killed.
* part_{inc|dec}_in_fligh() is implemented which automatically updates
part0 stats for parts other than part0.
* disk_map_sector_rcu() is updated to return part0 if no part matches.
Combined with the above changes, this makes NULL special case
handling in callers unnecessary.
* Separate stats show code paths for disk are collapsed into part
stats show code paths.
* Rename disk_stat_lock/unlock() to part_stat_lock/unlock()
While at it, reposition stat handling macros a bit and add missing
parentheses around macro parameters.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Till now, bdev->bd_part is set only if the bdev was for parts other
than part0. This patch makes bdev->bd_part always set so that code
paths don't have to differenciate common handling.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Implement {disk|part}_to_dev() and use them to access generic device
instead of directly dereferencing {disk|part}->dev. To make sure no
user is left behind, rename generic devices fields to __dev.
This is in preparation of unifying partition 0 handling with other
partitions.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
There are two variants of stat functions - ones prefixed with double
underbars which don't care about preemption and ones without which
disable preemption before manipulating per-cpu counters. It's unclear
whether the underbarred ones assume that preemtion is disabled on
entry as some callers don't do that.
This patch unifies diskstats access by implementing disk_stat_lock()
and disk_stat_unlock() which take care of both RCU (for partition
access) and preemption (for per-cpu counter access). diskstats access
should always be enclosed between the two functions. As such, there's
no need for the versions which disables preemption. They're removed
and double underbars ones are renamed to drop the underbars. As an
extra argument is added, there's no danger of using the old version
unconverted.
disk_stat_lock() uses get_cpu() and returns the cpu index and all
diskstat functions which access per-cpu counters now has @cpu
argument to help RT.
This change adds RCU or preemption operations at some places but also
collapses several preemption ops into one at others. Overall, the
performance difference should be negligible as all involved ops are
very lightweight per-cpu ones.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
* Implement disk_devt() and part_devt() and use them to directly
access devt instead of computing it from ->major and ->first_minor.
Note that all references to ->major and ->first_minor outside of
block layer is used to determine devt of the disk (the part0) and as
->major and ->first_minor will continue to represent devt for the
disk, converting these users aren't strictly necessary. However,
convert them for consistency.
* Implement disk_max_parts() to avoid directly deferencing
genhd->minors.
* Update bdget_disk() such that it doesn't assume consecutive minor
space.
* Move devt computation from register_disk() to add_disk() and make it
the only one (all other usages use the initially determined value).
These changes clean up the code and will help disk->part dereference
fix and extended block device numbers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
raid5 can overflow with more than 255 stripes, and we can increase it
to an int for free on both 32 and 64-bit archs due to the padding.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Remove hw_segments field from struct bio and struct request. Without virtual
merge accounting they have no purpose.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Moving the path activation to workqueue along with scsi_dh patches introduced
a race. It is due to the fact that the current_pgpath (in the multipath data
structure) can be modified if changes happen in any of the paths leading to
the lun. If the changes lead to current_pgpath being set to NULL, then it
leads to the invalid access which results in the panic below.
This patch fixes that by storing the pgpath to activate in the multipath data
structure and properly protecting it.
Note that if activate_path is called twice in succession with different pgpath,
with the second one being called before the first one is done, then activate
path will be called twice for the second pgpath, which is fine.
Unable to handle kernel paging request for data at address 0x00000020
Faulting instruction address: 0xd000000000aa1844
cpu 0x1: Vector: 300 (Data Access) at [c00000006b987a80]
pc: d000000000aa1844: .activate_path+0x30/0x218 [dm_multipath]
lr: c000000000087a2c: .run_workqueue+0x114/0x204
sp: c00000006b987d00
msr: 8000000000009032
dar: 20
dsisr: 40000000
current = 0xc0000000676bb3f0
paca = 0xc0000000006f3680
pid = 2528, comm = kmpath_handlerd
enter ? for help
[c00000006b987da0] c000000000087a2c .run_workqueue+0x114/0x204
[c00000006b987e40] c000000000088b58 .worker_thread+0x120/0x144
[c00000006b987f00] c00000000008ca70 .kthread+0x78/0xc4
[c00000006b987f90] c000000000027cc8 .kernel_thread+0x4c/0x68
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
If for any reason dm_merge_bvec() is given an offset beyond the end of the
device, avoid an oops and always allow one page to be added to an empty bio.
We'll reject the I/O later after the bio is submitted.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Some callers assume they can always add at least one page to an empty bio,
so dm_merge_bvec should not return 0 in this case: we'll reject the I/O
later after the bio is submitted.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
When two md arrays share some block device (e.g each uses different
partitions on the one device), a resync of one array will wait for
the resync on the other to finish.
This can be a long time and as it currently waits TASK_UNINTERRUPTIBLE,
the softlockup code notices and complains.
So use TASK_INTERRUPTIBLE instead and make sure to flush signals
before calling schedule.
Signed-off-by: NeilBrown <neilb@suse.de>
A recent patch to protect the rdev list with rcu locking leaves us
with a problem because we can sleep on memalloc while holding the
rcu lock.
The rcu lock is only needed while walking the linked list as
uninteresting devices (failed or spares) can be removed at any time.
So only take the rcu lock while actually walking the linked list.
Take a refcount on the rdev during the time when we drop the lock
and do the memalloc to start IO.
When we return to the locked code, all the interesting devices
on the list will not have moved, so we can simply use
list_for_each_continue_rcu to pick up where we left off.
Signed-off-by: NeilBrown <neilb@suse.de>
When stopping an md array, or just switching to read-only, we
currently call invalidate_partition while holding the mddev lock.
The main reason for this is probably to ensure all dirty buffers
are flushed (invalidate_partition calls fsync_bdev).
However if any dirty buffers are found, it will almost certainly cause
a deadlock as starting writeout will require an update to the
superblock, and performing that updates requires taking the mddev
lock - which is already held.
This deadlock can be demonstrated by running "reboot -f -n" with
a root filesystem on md/raid, and some dirty buffers in memory.
All other calls to stop an array should already happen after a flush.
The normal sequence is to stop using the array (e.g. umount) which
will cause __blkdev_put to call sync_blockdev. Then open the
array and issue the STOP_ARRAY ioctl while the buffers are all still
clean.
So this invalidate_partition is normally a no-op, except for one case
where it will cause a deadlock.
So remove it.
This patch possibly addresses the regression recored in
http://bugzilla.kernel.org/show_bug.cgi?id=11460
and
http://bugzilla.kernel.org/show_bug.cgi?id=11452
though it isn't yet clear how it ever worked.
Signed-off-by: NeilBrown <neilb@suse.de>
If a 'repair' is requested when an array is in a position to 'recover' raid1
will perform the repair while md believes a recovery is happening. Address
this at both ends, i.e. cancel check/repair requests upon detecting a
recover condition and do not call ->spare_active after completing a
check/repair.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The raid10 resync/recovery code currently limits the amount of
in-flight resync IO to 2Meg. This was copied from raid1 where
it seems quite adequate. However for raid10, some layouts require
a bit of seeking to perform a resync, and allowing a larger buffer
size means that the seeking can be significantly reduced.
There is probably no real need to limit the amount of in-flight
IO at all. Any shortage of memory will naturally reduce the
amount of buffer space available down to a set minimum, and any
concurrent normal IO will quickly cause resync IO to back off.
The only problem would be that normal IO has to wait for all resync IO
to finish, so a very large amount of resync IO could cause unpleasant
latency when normal IO starts up.
So: increase RESYNC_DEPTH to allow 32Meg of buffer (if memory is
available) which seems to be a good amount. Also reduce the amount
of memory reserved as there is no need to keep 2Meg just for resync if
memory is tight.
Thanks to Keld for the suggestion.
Cc: Keld Jørn Simonsen <keld@dkuug.dk>
Signed-off-by: NeilBrown <neilb@suse.de>
Removing faulty devices from an array is a two stage process.
First the device is moved from being a part of the active array
to being similar to a spare device. Then it can be removed
by a request from user space.
The first step is currently not performed for read-only arrays,
so the second step can never succeed.
So allow readonly arrays to remove failed devices (which aren't
blocked).
Signed-off-by: NeilBrown <neilb@suse.de>
When we have externally managed metadata, we need to mark a failed
device as 'Blocked' and not allow any writes until that device
have been marked as faulty in the metadata and the Blocked flag has
been removed.
However it is perfectly OK to allow read requests when there is a
Blocked device, and with a readonly array, there may not be any
metadata-handler watching for blocked devices.
So in raid5/raid6 only allow a Blocked device to interfere with
Write request or resync. Read requests go through untouched.
raid1 and raid10 already differentiate between read and write
properly.
Signed-off-by: NeilBrown <neilb@suse.de>
We cannot currently change the size of a write-intent bitmap.
So if we change the size of an array which has such a bitmap, it
tries to set bits beyond the end of the bitmap.
For now, simply reject any request to change the size of an array
which has a bitmap. mdadm can remove the bitmap and add a new one
after the array has changed size.
Signed-off-by: NeilBrown <neilb@suse.de>
A recent patch allowed do_md_stop to know whether it was being called
via an ioctl or not, and thus where to allow for an extra open file
descriptor when checking if it is in use.
This broke then switch to readonly performed by the shutdown notifier,
which needs to work even when the array is still (apparently) active
(as md doesn't get told when the filesystem becomes readonly).
So restore this feature by pretending that there can be lots of
file descriptors open, but we still want do_md_stop to switch to
readonly.
Signed-off-by: NeilBrown <neilb@suse.de>
If we reduce the 'safe_mode_delay', it could still wait for the old
delay to completely expire before doing anything about safe_mode.
Thus the effect if the change is delayed.
To make the effect more immediate, run the timeout function
immediately if the delay was reduced. This may cause it to run
slightly earlier that required, but that is the safer option.
Signed-off-by: NeilBrown <neilb@suse.de>
* 'for-linus' of git://neil.brown.name/md:
md: raid10: wake up frozen array
md: do not count blocked devices as spares
md: do not progress the resync process if the stripe was blocked
md: delay notification of 'active_idle' to the recovery thread
md: fix merge error
md: move async_tx_issue_pending_all outside spin_lock_irq
* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
md: the bitmap code needs to use blk_plug_device_unlocked()
block: add a blk_plug_device_unlocked() that grabs the queue lock
When rescheduling a bio in raid10, we wake up
the md thread, but if the array is frozen, this
will have no effect. This causes the array to
remain frozen for eternity. We add a wake_up
to allow the array to de-freeze. This code is
nearly identical to the raid1 code, which has
this fix already.
Signed-off-by: Arthur Jones <ajones@riverbed.com>
Signed-off-by: NeilBrown <neilb@suse.de>
remove_and_add_spares() assumes that failed devices have been hot-removed
from the array. Removal is skipped in the 'blocked' case so do not count a
device in this state as 'spare'.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
handle_stripe will take no action on a stripe when waiting for userspace
to unblock the array, so do not report completed sectors.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
multipath keeps a separate device table which may be
more current than the built-in one.
So we should make sure to always call ->attach whenever
a multipath map with hardware handler is instantiated.
And we should call ->detach on removal, too.
[sekharan: update as per comments from agk]
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The original STRIPE_OP_IO removal patch had the following hunk:
- for (i = conf->raid_disks; i--; ) {
+ for (i = conf->raid_disks; i--; )
set_bit(R5_Wantwrite, &sh->dev[i].flags);
- if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending))
- sh->ops.count++;
- }
However it appears the hunk became broken after merging:
- for (i = conf->raid_disks; i--; ) {
+ for (i = conf->raid_disks; i--; )
set_bit(R5_Wantwrite, &sh->dev[i].flags);
set_bit(R5_LOCKED, &dev->flags);
s.locked++;
- if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending))
- sh->ops.count++;
- }
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Some dma drivers need to call spin_lock_bh in their device_issue_pending
routines. This change avoids:
WARNING: at kernel/softirq.c:136 local_bh_enable_ip+0x3a/0x85()
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
* 'for-linus' of git://neil.brown.name/md: (52 commits)
md: Protect access to mddev->disks list using RCU
md: only count actual openers as access which prevent a 'stop'
md: linear: Make array_size sector-based and rename it to array_sectors.
md: Make mddev->array_size sector-based.
md: Make super_type->rdev_size_change() take sector-based sizes.
md: Fix check for overlapping devices.
md: Tidy up rdev_size_store a bit:
md: Remove some unused macros.
md: Turn rdev->sb_offset into a sector-based quantity.
md: Make calc_dev_sboffset() return a sector count.
md: Replace calc_dev_size() by calc_num_sectors().
md: Make update_size() take the number of sectors.
md: Better control of when do_md_stop is allowed to stop the array.
md: get_disk_info(): Don't convert between signed and unsigned and back.
md: Simplify restart_array().
md: alloc_disk_sb(): Return proper error value.
md: Simplify sb_equal().
md: Simplify uuid_equal().
md: sb_equal(): Fix misleading printk.
md: Fix a typo in the comment to cmd_match().
...
This patch implements biovec merge function for crypt target.
If the underlying device has merge function defined, call it.
If not, keep precomputed value.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Remove max_sector restriction - merge function replaced it.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
This patch implements biovec merge function for linear target.
If the underlying device has merge function defined, call it.
If not, keep precomputed value.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Introduce a bvec merge function for device mapper devices
for dynamic size restrictions.
This code ensures the requested biovec lies within a single
target and then calls a target-specific function to check
against any constraints imposed by underlying devices.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Change snapshot per-module mempool to per-device mempool.
Per-module mempools could cause a deadlock if multiple
snapshot devices are stacked above each other.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Fix a race condition that returns incorrect data when a write causes an
exception to be allocated whilst a read is still in flight.
The race condition happens as follows:
* A read to non-reallocated sector in the snapshot is submitted so that the
read is routed to the original device.
* A write to the original device is submitted. The write causes an exception
that reallocates the block. The write proceeds.
* The original read is dequeued and reads the wrong data.
This race can be triggered with CFQ scheduler and one thread writing and
multiple threads reading simultaneously.
(This patch relies upon the earlier dm-kcopyd-per-device.patch to avoid a
deadlock.)
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Whenever a snapshot read gets mapped through to the origin, track it in
a per-snapshot hash table indexed by chunk number, using memory allocated
from a new per-snapshot mempool.
We need to track these reads to avoid race conditions which will be fixed
by patches that follow.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Return a specific error message if there are an invalid number of multipath
arguments.
This invalid command returns an "Unknown error" because the ti->error field is
not set
dmsetup create --table '0 2 multipath 0 0 1 1 round-robin 0 1 1 /dev/sdh' mpath0
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
dm_dirty_log_{init,exit}() can now become static.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Free path selector if the arguments are invalid.
This command (note that it is invalid) causes reference leak on module
"dm_round_robin" and prevents the module from being removed.
dmsetup create --table '0 2 multipath 0 0 1 1 round-robin /dev/sdh' mpath0
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
All modifications and most access to the mddev->disks list are made
under the reconfig_mutex lock. However there are three places where
the list is walked without any locking. If a reconfig happens at this
time, havoc (and oops) can ensue.
So use RCU to protect these accesses:
- wrap them in rcu_read_{,un}lock()
- use list_for_each_entry_rcu
- add to the list with list_add_rcu
- delete from the list with list_del_rcu
- delay the 'free' with call_rcu rather than schedule_work
Note that export_rdev did a list_del_init on this list. In almost all
cases the entry was not in the list anymore so it was a no-op and so
safe. It is no longer safe as after list_del_rcu we may not touch
the list_head.
An audit shows that export_rdev is called:
- after unbind_rdev_from_array, in which case the delete has
already been done,
- after bind_rdev_to_array fails, in which case the delete isn't needed.
- before the device has been put on a list at all (e.g. in
add_new_disk where reading the superblock fails).
- and in autorun devices after a failure when the device is on a
different list.
So remove the list_del_init call from export_rdev, and add it back
immediately before the called to export_rdev for that last case.
Note also that ->same_set is sometimes used for lists other than
mddev->list (e.g. candidates). In these cases rcu is not needed.
Signed-off-by: NeilBrown <neilb@suse.de>
Open isn't the only thing that increments ->active. e.g. reading
/proc/mdstat will increment it briefly. So to avoid false positives
in testing for concurrent access, introduce a new counter that counts
just the number of times the md device it open.
Signed-off-by: NeilBrown <neilb@suse.de>
This patch renames the array_size field of struct mddev_s to array_sectors
and converts all instances to use units of 512 byte sectors instead of 1k
blocks.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: NeilBrown <neilb@suse.de>
Also, change the type of the size parameter from unsigned long long to
sector_t and rename it to num_sectors.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: NeilBrown <neilb@suse.de>
The checks in overlaps() expect all parameters either in block-based
or sector-based quantities. However, its single caller passes two
rdev->data_offset arguments as well as two rdev->size arguments, the
former being sector counts while the latter are measured in 1K blocks.
This could cause rdev_size_store() to accept an invalid size from user
space. Fix it by passing only sector-based quantities to overlaps().
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: NeilBrown <neilb@suse.de>
- used strict_strtoull in place of simple_strtoull
- use my_mddev in place of rdev->mddev (they have the same value)
and more significantly,
- don't adjust mddev->size to fit, rather reject changes which make
rdev->size smaller than mddev->size
Adjusting mddev->size is a hangover from bind_rdev_to_array which
does a similar thing. But it really is a better design to insist that
mddev->size is set as required, then the rdev->sizes are set to allow
for that. The previous way invites confusion.
Signed-off-by: NeilBrown <neilb@suse.de>
* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6: (102 commits)
[SCSI] scsi_dh: fix kconfig related build errors
[SCSI] sym53c8xx: Fix bogus sym_que_entry re-implementation of container_of
[SCSI] scsi_cmnd.h: remove double inclusion of linux/blkdev.h
[SCSI] make struct scsi_{host,target}_type static
[SCSI] fix locking in host use of blk_plug_device()
[SCSI] zfcp: Cleanup external header file
[SCSI] zfcp: Cleanup code in zfcp_erp.c
[SCSI] zfcp: zfcp_fsf cleanup.
[SCSI] zfcp: consolidate sysfs things into one file.
[SCSI] zfcp: Cleanup of code in zfcp_aux.c
[SCSI] zfcp: Cleanup of code in zfcp_scsi.c
[SCSI] zfcp: Move status accessors from zfcp to SCSI include file.
[SCSI] zfcp: Small QDIO cleanups
[SCSI] zfcp: Adapter reopen for large number of unsolicited status
[SCSI] zfcp: Fix error checking for ELS ADISC requests
[SCSI] zfcp: wait until adapter is finished with ERP during auto-port
[SCSI] ibmvfc: IBM Power Virtual Fibre Channel Adapter Client Driver
[SCSI] sg: Add target reset support
[SCSI] lib: Add support for the T10 (SCSI) Data Integrity Field CRC
[SCSI] sd: Move scsi_disk() accessor function to sd.h
...
Do not automatically "select" SCSI_DH for dm-multipath. If SCSI_DH
doesn't exist,just do not allow hardware handlers to be used.
Handle SCSI_DH being a module also. Make sure it doesn't allow DM_MULTIPATH
to be compiled in when SCSI_DH is a module.
[jejb: added comment for Kconfig syntax]
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
* 'for-linus' of git://git.kernel.dk/linux-2.6-block: (37 commits)
splice: fix generic_file_splice_read() race with page invalidation
ramfs: enable splice write
drivers/block/pktcdvd.c: avoid useless memset
cdrom: revert commit 22a9189 (cdrom: use kmalloced buffers instead of buffers on stack)
scsi: sr avoids useless buffer allocation
block: blk_rq_map_kern uses the bounce buffers for stack buffers
block: add blk_queue_update_dma_pad
DAC960: push down BKL
pktcdvd: push BKL down into driver
paride: push ioctl down into driver
block: use get_unaligned_* helpers
block: extend queue_flag bitops
block: request_module(): use format string
Add bvec_merge_data to handle stacked devices and ->merge_bvec()
block: integrity flags can't use bit ops on unsigned short
cmdfilter: extend default read filter
sg: fix odd style (extra parenthesis) introduced by cmd filter patch
block: add bounce support to blk_rq_map_user_iov
cfq-iosched: get rid of enable_idle being unused warning
allow userspace to modify scsi command filter on per device basis
...
Rename it to sb_start to make sure all users have been converted.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
As BLOCK_SIZE_BITS is 10 and
MD_NEW_SIZE_SECTORS(2 * x) = 2 * NEW_SIZE_BLOCKS(x),
the return value of calc_dev_sboffset() doubles. Fix up all three
callers accordingly.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
Number of sectors is the preferred unit for sizes of raid devices,
so change calc_dev_size() so that it returns this unit instead of
the number of 1K blocks.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
Changing the internal representations of sizes of raid devices
from 1K blocks to sector counts (512B units) is desirable because
it allows to get rid of many divisions/multiplications and unnecessary
casts that are present in the current code.
This patch is a first step in this direction. It replaces the old
1K-based "size" argument of update_size() by "num_sectors" and
fixes up its two callers.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
do_md_stop check the number of active users before allowing the array
to be stopped.
Two problems:
1/ it assumes the request is coming through an open file descriptor
(via ioctl) so it allows for that. This is not always the case.
2/ it doesn't do the check it the array hasn't been activated.
This is not good for cases when we use an inactive array to hold
some devices in a container.
Signed-off-by: Neil Brown <neilb@suse.de>
The current code copies a signed int from user space, converts it to
unsigned and passes the unsigned value to find_rdev_nr() which expects
a signed value. Simply pass the signed value from user space directly.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
If alloc_page() fails, ENOMEM is a more suitable error value
than EINVAL.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
The only caller of sb_equal() tests the return value against
zero, so it's OK to return the negated return value of memcmp().
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
Remove the dubious attempt to prefer 'compute' over 'read'. Not only is it
wrong given commit c337869d (md: do not compute parity unless it is on a failed
drive), but it can trigger a BUG_ON in handle_parity_checks5().
Cc: <stable@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
- Remove superfluous parentheses.
- Make format string match the type of the variable that is printed.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
In case pers->run() succeeds but creating the bitmap fails, we
print an error message stating that pers->run() has failed.
Print this message only if pers->run() really failed.
Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Neil Brown <neilb@suse.de>
When devices are stacked, one device's merge_bvec_fn may need to perform
the mapping and then call one or more functions for its underlying devices.
The following bio fields are used:
bio->bi_sector
bio->bi_bdev
bio->bi_size
bio->bi_rw using bio_data_dir()
This patch creates a new struct bvec_merge_data holding a copy of those
fields to avoid having to change them directly in the struct bio when
going down the stack only to have to change them back again on the way
back up. (And then when the bio gets mapped for real, the whole
exercise gets repeated, but that's a problem for another day...)
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Cc: Neil Brown <neilb@suse.de>
Cc: Milan Broz <mbroz@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Add cond_resched() to prevent monopolising CPU when processing large bios.
dm-crypt processes encryption of bios in sector units. If the bio request
is big it can spend a long time in the encryption call.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Tested-by: Yan Li <elliot.li.tech@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
md_allow_write() marks the metadata dirty while holding mddev->lock and then
waits for the write to complete. For externally managed metadata this causes a
deadlock as userspace needs to take the lock to communicate that the metadata
update has completed.
Change md_allow_write() in the 'external' case to start the 'mark active'
operation and then return -EAGAIN. The expected side effects while waiting for
userspace to write 'active' to 'array_state' are holding off reshape (code
currently handles -ENOMEM), cause some 'stripe_cache_size' change requests to
fail, cause some GET_BITMAP_FILE ioctl requests to fall back to GFP_NOIO, and
cause updates to 'raid_disks' to fail. Except for 'stripe_cache_size' changes
these failures can be mitigated by coordinating with mdmon.
md_write_start() still prevents writes from occurring until the metadata
handler has had a chance to take action as it unconditionally waits for
MD_CHANGE_CLEAN to be cleared.
[neilb@suse.de: return -EAGAIN, try GFP_NOIO]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
From: Dan Williams <dan.j.williams@intel.com>
Commit a4456856 refactored some of the deep code paths in raid5.c into separate
functions. The names chosen at the time do not consistently indicate what is
going to happen to the stripe. So, update the names, and since a stripe is a
cache element use cache semantics like fill, dirty, and clean.
(also, fix up the indentation in fetch_block5)
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
Neil said:
> At the end of ops_run_compute5 you have:
> /* ack now if postxor is not set to be run */
> if (tx && !test_bit(STRIPE_OP_POSTXOR, &s->ops_run))
> async_tx_ack(tx);
>
> It looks odd having that test there. Would it fit in raid5_run_ops
> better?
The intended global interpretation is that raid5_run_ops can build a chain
of xor and memcpy operations. When MD registers the compute-xor it tells
async_tx to keep the operation handle around so that another item in the
dependency chain can be submitted. If we are just computing a block to
satisfy a read then we can terminate the chain immediately. raid5_run_ops
gives a better context for this test since it cares about the entire chain.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
Currently ops_run_biodrain and other locations have extra logic to determine
which blocks are processed in the prexor and non-prexor cases. This can be
eliminated if handle_write_operations5 flags the blocks to be processed in all
cases via R5_Wantdrain. The presence of the prexor operation is tracked in
sh->reconstruct_state.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
Track the state of reconstruct operations (recalculating the parity block
usually due to incoming writes, or as part of array expansion) Reduces the
scope of the STRIPE_OP_{BIODRAIN,PREXOR,POSTXOR} flags to only tracking whether
a reconstruct operation has been requested via the ops_request field of struct
stripe_head_state.
This is the final step in the removal of ops.{pending,ack,complete,count}, i.e.
the STRIPE_OP_{BIODRAIN,PREXOR,POSTXOR} flags only request an operation and do
not track the state of the operation.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
Track the state of compute operations (recalculating a block from all the other
blocks in a stripe) with a state flag. Reduces the scope of the
STRIPE_OP_COMPUTE_BLK flag to only tracking whether a compute operation has
been requested via the ops_request field of struct stripe_head_state.
Note, the compute operation that is performed in the course of doing a 'repair'
operation (check the parity block, recalculate it and write it back if the
check result is not zero) is tracked separately with the 'check_state'
variable. Compute operations are held off while a 'check' is in progress, and
moving this check out to handle_issuing_new_read_requests5 the helper routine
__handle_issuing_new_read_requests5 can be simplified.
This is another step towards the removal of ops.{pending,ack,complete,count},
i.e. STRIPE_OP_COMPUTE_BLK only requests an operation and does not track the
state of the operation.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
Track the state of read operations (copying data from the stripe cache to bio
buffers outside the lock) with a state flag. Reduce the scope of the
STRIPE_OP_BIOFILL flag to only tracking whether a biofill operation has been
requested via the ops_request field of struct stripe_head_state.
This is another step towards the removal of ops.{pending,ack,complete,count},
i.e. STRIPE_OP_BIOFILL only requests an operation and does not track the state
of the operation.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
The STRIPE_OP_* flags record the state of stripe operations which are
performed outside the stripe lock. Their use in indicating which
operations need to be run is straightforward; however, interpolating what
the next state of the stripe should be based on a given combination of
these flags is not straightforward, and has led to bugs. An easier to read
implementation with minimal degrees of freedom is needed.
Towards this goal, this patch introduces explicit states to replace what was
previously interpolated from the STRIPE_OP_* flags. For now this only converts
the handle_parity_checks5 path, removing a user of the
ops.{pending,ack,complete,count} fields of struct stripe_operations.
This conversion also found a remaining issue with the current code. There is
a small window for a drive to fail between when we schedule a repair and when
the parity calculation for that repair completes. When this happens we will
writeback to 'failed_num' when we really want to write back to 'pd_idx'.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
Let the raid6 path call ops_run_io to get pending i/o submitted.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
In handle_stripe after taking sh->lock we sample some bits into 's' (struct
stripe_head_state):
s.syncing = test_bit(STRIPE_SYNCING, &sh->state);
s.expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
s.expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
Use these values from 's' in ops_run_io() rather than re-sampling the bits.
This ensures a consistent snapshot (as seen under sh->lock) is used.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
The R5_Want{Read,Write} flags already gate i/o. So, this flag is
superfluous and we can unconditionally call ops_run_io().
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Dan Williams <dan.j.williams@intel.com>
This micro-optimization allowed the raid code to skip a re-read of the
parity block after checking parity. It took advantage of the fact that
xor-offload-engines have their own internal result buffer and can check
parity without writing to memory. Remove it for the following reasons:
1/ It is a layering violation for MD to need to manage the DMA and
non-DMA paths within async_xor_zero_sum
2/ Bad precedent to toggle the 'ops' flags outside the lock
3/ Hard to realize a performance gain as reads will not need an updated
parity block and writes will dirty it anyways.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
From: Chris Webb <chris@arachsys.com>
Allow /sys/block/mdX/md/rdY/size to change on running arrays, moving the
superblock if necessary for this metadata version. We prevent the available
space from shrinking to less than the used size, and allow it to be set to zero
to fill all the available space on the underlying device.
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: Neil Brown <neilb@suse.de>
The important state change happens during an interrupt
in md_error. So just set a flag there and call sysfs_notify
later in process context.
Signed-off-by: Neil Brown <neilb@suse.de>
