When we get any IO error during a recovery (rebuilding a spare), we abort
the recovery and restart it.
For RAID6 (and multi-drive RAID1) it may not be best to restart at the
beginning: when multiple failures can be tolerated, the recovery may be
able to continue and re-doing all that has already been done doesn't make
sense.
We already have the infrastructure to record where a recovery is up to
and restart from there, but it is not being used properly.
This is because:
- We sometimes abort with MD_RECOVERY_ERR rather than just MD_RECOVERY_INTR,
which causes the recovery not be be checkpointed.
- We remove spares and then re-added them which loses important state
information.
The distinction between MD_RECOVERY_ERR and MD_RECOVERY_INTR really isn't
needed. If there is an error, the relevant drive will be marked as
Faulty, and that is enough to ensure correct handling of the error. So we
first remove MD_RECOVERY_ERR, changing some of the uses of it to
MD_RECOVERY_INTR.
Then we cause the attempt to remove a non-faulty device from an array to
fail (unless recovery is impossible as the array is too degraded). Then
when remove_and_add_spares attempts to remove the devices on which
recovery can continue, it will fail, they will remain in place, and
recovery will continue on them as desired.
Issue: If we are halfway through rebuilding a spare and another drive
fails, and a new spare is immediately available, do we want to:
1/ complete the current rebuild, then go back and rebuild the new spare or
2/ restart the rebuild from the start and rebuild both devices in
parallel.
Both options can be argued for. The code currently takes option 2 as
a/ this requires least code change
b/ this results in a minimally-degraded array in minimal time.
Cc: "Eivind Sarto" <ivan@kasenna.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In some configurations, a raid6 resync can be limited by CPU speed
(Calculating P and Q and moving data) rather than by device speed. In
these cases there is nothing to be gained byt serialising resync of arrays
that share a device, and doing the resync in parallel can provide benefit.
So add a sysfs tunable to flag an array as being allowed to resync in
parallel with other arrays that use (a different part of) the same device.
Signed-off-by: Bernd Schubert <bs@q-leap.de>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The atomic_t type is 32bit but a 64bit system can have more than 2^32
pages of virtual address space available. Without this we overflow on
ludicrously large mappings
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
<linux/types.h> can't be used together with <sys/ustat.h> because they
both define struct ustat:
$ cat test.c
#include <sys/ustat.h>
#include <linux/types.h>
$ gcc -c test.c
In file included from test.c:2:
/usr/include/linux/types.h:165: error: redefinition of 'struct ustat'
has been reported a while ago to debian, but seems to have been
lost in cat fighting: http://bugs.debian.org/429064
Signed-off-by: maximilian attems <max@stro.at>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Compiling ppc64_defconfig with gcc 4.3 gives thes warnings:
arch/powerpc/sysdev/mpic.c: In function 'mpic_irq_get_priority':
arch/powerpc/sysdev/mpic.c:1351: warning: 'is_ipi' may be used uninitialized in this function
arch/powerpc/sysdev/mpic.c: In function 'mpic_irq_set_priority':
arch/powerpc/sysdev/mpic.c:1328: warning: 'is_ipi' may be used uninitialized in this function
It turns out that in the cases where is_ipi is uninitialized, another
variable (mpic) will be NULL and it is dereferenced. Protect against
this by returning if mpic is NULL in mpic_irq_set_priority, and removing
mpic_irq_get_priority completely as it has no in tree callers.
This has the nice side effect of making the warning go away.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
net: The world is not perfect patch.
tcp: Make prior_ssthresh a u32
xfrm_user: Remove zero length key checks.
net/ipv4/arp.c: Use common hex_asc helpers
cassini: Only use chip checksum for ipv4 packets.
tcp: TCP connection times out if ICMP frag needed is delayed
netfilter: Move linux/types.h inclusions outside of #ifdef __KERNEL__
af_key: Fix selector family initialization.
libertas: Fix ethtool statistics
mac80211: fix NULL pointer dereference in ieee80211_compatible_rates
mac80211: don't claim iwspy support
orinoco_cs: add ID for SpeedStream wireless adapters
hostap_cs: add ID for Conceptronic CON11CPro
rtl8187: resource leak in error case
ath5k: Fix loop variable initializations
If previous window was above representable values of u16,
strange things will happen if undo with the truncated value
is called for. Alternatively, this could be fixed by some
max trickery but that would limit undoing high-speed undos.
Adds 16-bit hole but there isn't anything to fill it with.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Greg Steuck <greg@nest.cx> points out that some of the netfilter
headers can't be used in userspace without including linux/types.h
first. The headers include their own linux/types.h include statements,
these are stripped by make headers-install because they are inside
#ifdef __KERNEL__ however. Move them out to fix this.
Reported and Tested by Greg Steuck.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'for-2.6.26' of git://linux-nfs.org/~bfields/linux: (25 commits)
svcrdma: Verify read-list fits within RPCSVC_MAXPAGES
svcrdma: Change svc_rdma_send_error return type to void
svcrdma: Copy transport address and arm CQ before calling rdma_accept
svcrdma: Set rqstp transport address in rdma_read_complete function
svcrdma: Use ib verbs version of dma_unmap
svcrdma: Cleanup queued, but unprocessed I/O in svc_rdma_free
svcrdma: Move the QP and cm_id destruction to svc_rdma_free
svcrdma: Add reference for each SQ/RQ WR
svcrdma: Move destroy to kernel thread
svcrdma: Shrink scope of spinlock on RQ CQ
svcrdma: Use standard Linux lists for context cache
svcrdma: Simplify RDMA_READ deferral buffer management
svcrdma: Remove unused READ_DONE context flags bit
svcrdma: Return error from rdma_read_xdr so caller knows to free context
svcrdma: Fix error handling during listening endpoint creation
svcrdma: Free context on post_recv error in send_reply
svcrdma: Free context on ib_post_recv error
svcrdma: Add put of connection ESTABLISHED reference in rdma_cma_handler
svcrdma: Fix return value in svc_rdma_send
svcrdma: Fix race with dto_tasklet in svc_rdma_send
...
* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
USB: CDC WDM driver
USB: ehci-orion: the Orion EHCI root hub does have a Transaction Translator
USB: serial: ch341: New VID/PID for CH341 USB-serial
USB: build fix
USB: pxa27x_udc - Fix Oops
USB: OPTION: fix name of Onda MSA501HS HSDPA modem
USB: add TELIT HDSPA UC864-E modem to option driver
usb-serial: Use ftdi_sio driver for RATOC REX-USB60F
Propagate feature bits from the NETDEV_FEAT_CHANGE notifier. For now
only TSO is propagated for devices that announce their ability to
support TSO in combination with VLAN accel by setting the NETIF_F_VLAN_TSO
flag.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
We want to have the drvdata field set properly when creating the device
as sysfs callbacks can assume it is present and it can race the later
setting of this field.
So, create two new functions, deviec_create_vargs() and
device_create_drvdata() that take this new field.
device_create_drvdata() will go away in 2.6.27 as the drvdata field will
just be moved to the device_create() call as it should be.
Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
vidiocgmbuf() does this:
mutex_lock(&fh->cap.vb_lock);
retval = videobuf_mmap_setup(&fh->cap, gbuffers, gbufsize,
V4L2_MEMORY_MMAP);
and videobuf_mmap_setup() then just does
mutex_lock(&q->vb_lock);
ret = __videobuf_mmap_setup(q, bcount, bsize, memory);
mutex_unlock(&q->vb_lock);
which is an obvious double-take deadlock.
This patch fixes this by having vidiocgmbuf() just call the
__videobuf_mmap_setup function instead.
Acked-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Reported-by: Koos Vriezen <koos.vriezen@gmail.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The x86_64 pgd_bad(), pud_bad(), pmd_bad() inlines have differed from
their x86_32 counterparts in a couple of ways: they've been unnecessarily
weak (e.g. letting 0 or 1 count as good), and were typed as unsigned long.
Strengthen them and return int.
The PAE pmd_bad was too weak before, allowing any junk in the upper half;
but got strengthened by the patch correcting its ~PAGE_MASK to ~PTE_MASK.
The PAE pud_bad already said ~PTE_MASK; and since it folds into pgd_bad,
and we don't set the protection bits at that level, it'll do as is.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
