__try_to_free_cp_buf(), __process_buffer(), and __wait_cp_io() test
BH_Uptodate flag to detect write I/O errors on metadata buffers. But by
commit 95450f5a7e53d5752ce1a0d0b8282e10fe745ae0 "ext3: don't read inode
block if the buffer has a write error"(*), BH_Uptodate flag can be set to
inode buffers with BH_Write_EIO in order to avoid reading old inode data.
So now, we have to test BH_Write_EIO flag of checkpointing inode buffers
instead of BH_Uptodate. This patch does it.
Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com> Acked-by: Jan Kara <jack@suse.cz> Acked-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When a checkpointing IO fails, current JBD code doesn't check the error
and continue journaling. This means latest metadata can be lost from both
the journal and filesystem.
This patch leaves the failed metadata blocks in the journal space and
aborts journaling in the case of log_do_checkpoint(). To achieve this, we
need to do:
1. don't remove the failed buffer from the checkpoint list where in
the case of __try_to_free_cp_buf() because it may be released or
overwritten by a later transaction
2. log_do_checkpoint() is the last chance, remove the failed buffer
from the checkpoint list and abort the journal
3. when checkpointing fails, don't update the journal super block to
prevent the journaled contents from being cleaned. For safety,
don't update j_tail and j_tail_sequence either
4. when checkpointing fails, notify this error to the ext3 layer so
that ext3 don't clear the needs_recovery flag, otherwise the
journaled contents are ignored and cleaned in the recovery phase
5. if the recovery fails, keep the needs_recovery flag
6. prevent cleanup_journal_tail() from being called between
__journal_drop_transaction() and journal_abort() (a race issue
between journal_flush() and __log_wait_for_space()
Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com> Acked-by: Jan Kara <jack@suse.cz> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some Apple boxes evidently require us to set SCI_EN on resume
directly, because if we don't do that, they hung somewhere in the
resume code path. Moreover, on these boxes it is not sufficient to
use acpi_enable() to turn ACPI on during resume. All of this is
against the ACPI specification which states that (1) the BIOS is
supposed to return from the S3 sleep state with ACPI enabled
(SCI_EN set) and (2) the SCI_EN bit is owned by the hardware and we
are not supposed to change it.
For this reason, blacklist the affected systems so that the SCI_EN
bit is set during resume on them.
[NOTE: Unconditional setting SCI_EN for all system on resume doesn't
work, because it makes some other systems crash (that's to be
expected). Also, it is not entirely clear right now if all of the
Apple boxes require this workaround.]
This patch fixes the recent regression tracked as
http://bugzilla.kernel.org/show_bug.cgi?id=12038
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Tested-by: Tino Keitel <tino.keitel@gmx.de> Tested-by: Bob Copeland <me@bobcopeland.com> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Changeset a238b790d5f99c7832f9b73ac8847025815b85f7 (Call fasync()
functions without the BKL) introduced a race which could leave
file->f_flags in a state inconsistent with what the underlying
driver/filesystem believes. Revert that change, and also fix the same
races in ioctl_fioasync() and ioctl_fionbio().
This is a minimal, short-term fix; the real fix will not involve the
BKL.
There's no point in having too short SG_IO timeouts, since if the
command does end up timing out, we'll end up through the reset sequence
that is several seconds long in order to abort the command that timed
out.
As a result, shorter timeouts than a few seconds simply do not make
sense, as the recovery would be longer than the timeout itself.
Add a BLK_MIN_SG_TIMEOUT to match the existign BLK_DEFAULT_SG_TIMEOUT.
Copy the FPU state to the task's thread_info->fpregs for the VIS emulation
functions to access.
Signed-off-by: Hong H. Pham <hong.pham@windriver.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch fixes some bugs in VIS emulation that cause the GCC test
failure
FAIL: gcc.target/sparc/pdist-3.c execution test
for both 32-bit and 64-bit testing on hardware lacking these
instructions. The emulation code for the pdist instruction uses
RS1(insn) for both source registers rs1 and rs2, which is obviously
wrong and leads to the instruction doing nothing (the observed
problem), and further inspection of the code shows that RS1 uses a
shift of 24 and RD a shift of 25, which clearly cannot both be right;
examining SPARC documentation indicates the correct shift for RS1 is
14.
This patch fixes the bug if single-stepping over the affected
instruction in the debugger, but not if the testcase is run
standalone. For that, Wind River has another patch I hope they will
send as a followup to this patch submission.
Signed-off-by: Joseph Myers <joseph@codesourcery.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
>The SPARC64 kernel code for PTRACE_SETFPREGS64 appears to be an exact copy
>of that for PTRACE_GETFPREGS64. This means that gdbserver and native
>64-bit GDB cannot set floating-point registers.
It looks like a simple typo.
Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is a problem discovered in recent versions of ATI Mach64 driver
in X.org on sparc64 architecture. In short, the driver fails to mmap
MMIO aperture (PCI resource #2).
I've found that kernel's __pci_mmap_make_offset() returns EINVAL. It
checks whether user attempts to mmap more than the resource length,
which is 0x1000 bytes in our case. But PAGE_SIZE on SPARC64 is 0x2000
and this is what actually is being mmaped. So __pci_mmap_make_offset()
failed for this PCI resource.
Signed-off-by: Max Dmitrichenko <dmitrmax@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alexander Beregalov reports oops in __bzero() called from
copy_from_user_fixup() called from iov_iter_copy_from_user_atomic(),
when running dbench on tmpfs on sparc64: its __copy_from_user_inatomic
and __copy_to_user_inatomic should be avoiding, not calling, the fixups.
Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Current UDP multicast delivery is not namespace aware.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com> Acked-by: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This fixes a TX hang reported by Jesper Dangaard Brouer.
When an architecutre cannot provide a fully functional
64-bit atomic readq/writeq, the driver must implement
it's own. This is because only the driver can say whether
doing something like using two 32-bit reads to implement
the full 64-bit read will actually work properly.
In particular one of the issues is whether the top 32-bits
or the bottom 32-bits of the 64-bit register should be read
first. There could be side effects, and in fact that is
exactly the problem here.
The TX_CS register has counters in the upper 32-bits and
state bits in the lower 32-bits. A read clears the state
bits.
We would read the counter half before the state bit half.
That first read would clear the state bits, and then the
driver thinks that no interrupts are pending because the
interrupt indication state bits are seen clear every time.
Fix this by reading the bottom half before the upper half.
Tested-by: Jesper Dangaard Brouer <jdb@comx.dk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently libata uses four methods to detect device presence.
1. PHY status if available.
2. TF register R/W test (only promotes presence, never demotes)
3. device signature after reset
4. IDENTIFY failure detection in SFF state machine
Combination of the above works well in most cases but recently there
have been a few reports where a phantom device causes unnecessary
delay during probe. In both cases, PHY status wasn't available. In
one case, it passed #2 and #3 and failed IDENTIFY with ATA_ERR which
didn't qualify as #4. The other failed #2 but as it passed #3 and #4,
it still caused failure.
In both cases, phantom device reported diagnostic failure, so these
cases can be safely worked around by considering any !ATA_DRQ IDENTIFY
failure as NODEV_HINT if diagnostic failure is set.
In ordered mode, if a file data buffer being dirtied exists in the
committing transaction, we write the buffer to the disk, move it from the
committing transaction to the running transaction, then dirty it. But we
don't have to remove the buffer from the committing transaction when the
buffer couldn't be written out, otherwise it would miss the error and the
committing transaction would not abort.
This patch adds an error check before removing the buffer from the
committing transaction.
Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com> Acked-by: Jan Kara <jack@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We could run into ENOSPC error on ext3, even when there is free blocks on
the filesystem.
The problem is triggered in the case the goal block group has 0 free
blocks , and the rest block groups are skipped due to the check of
"free_blocks < windowsz/2". Current code could fall back to non
reservation allocation to prevent early ENOSPC after examing all the block
groups with reservation on , but this code was bypassed if the reservation
window is turned off already, which is true in this case.
This patch fixed two issues:
1) We don't need to turn off block reservation if the goal block group has
0 free blocks left and continue search for the rest of block groups.
Current code the intention is to turn off the block reservation if the
goal allocation group has a few (some) free blocks left (not enough for
make the desired reservation window),to try to allocation in the goal
block group, to get better locality. But if the goal blocks have 0 free
blocks, it should leave the block reservation on, and continues search for
the next block groups,rather than turn off block reservation completely.
2) we don't need to check the window size if the block reservation is off.
The problem was originally found and fixed in ext4.
We could run into ENOSPC error on ext2, even when there is free blocks on
the filesystem.
The problem is triggered in the case the goal block group has 0 free
blocks , and the rest block groups are skipped due to the check of
"free_blocks < windowsz/2". Current code could fall back to non
reservation allocation to prevent early ENOSPC after examing all the block
groups with reservation on , but this code was bypassed if the reservation
window is turned off already, which is true in this case.
This patch fixed two issues:
1) We don't need to turn off block reservation if the goal block group has
0 free blocks left and continue search for the rest of block groups.
Current code the intention is to turn off the block reservation if the
goal allocation group has a few (some) free blocks left (not enough for
make the desired reservation window),to try to allocation in the goal
block group, to get better locality. But if the goal blocks have 0 free
blocks, it should leave the block reservation on, and continues search for
the next block groups,rather than turn off block reservation completely.
2) we don't need to check the window size if the block reservation is off.
The problem was originally found and fixed in ext4.
When trying to resize a ext3 fs and you run out of reserved gdt blocks,
you get an error that doesn't actually tell you what went wrong, it just
says that the gdb it picked is not correct, which is the case since you
don't have any reserved gdt blocks left. This patch adds a check to make
sure you have reserved gdt blocks to use, and if not prints out a more
relevant error.
Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: <linux-ext4@vger.kernel.org> Cc: Andreas Dilger <adilger@sun.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Willy Tarreau <w@1wt.eu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix a regression caused by commit 6a897cf4, "ext3: fix ext3_dx_readdir
hash collision handling", where deleting files in a large directory
(requiring more than one getdents system call), results in some
filenames being returned twice. This was caused by a failure to
update info->curr_hash and info->curr_minor_hash, so that if the
directory had gotten modified since the last getdents() system call
(as would be the case if the user is running "rm -r" or "git clean"),
a directory entry would get returned twice to the userspace.
This patch fixes the bug reported by Markus Trippelsdorf at:
http://bugzilla.kernel.org/show_bug.cgi?id=11844
When initializing an uninitialized block group in ext4_new_inode(),
its block group checksum must be re-calculated. This fixes a race
when several threads try to allocate a new inode in an UNINIT'd group.
There is some question whether we need to be initializing the block
bitmap in ext4_new_inode() at all, but for now, if we are going to
init the block group, let's eliminate the race.
This fixes a 2.6.27 regression which was introduced in commit a02908f1.
We weren't passing the chunk parameter down to the two subections,
ext4_indirect_trans_blocks() and ext4_ext_index_trans_blocks(), with
the result that massively overestimate the amount of credits needed by
ext4_da_writepages, especially in the non-extents case. This causes
failures especially on /boot partitions, which tend to be small and
non-extent using since GRUB doesn't handle extents.
This patch fixes the bug reported by Joseph Fannin at:
http://bugzilla.kernel.org/show_bug.cgi?id=11964
In ext4_sync_fs, we only wait for a commit to finish if we started it,
but there may be one already in progress which will not be synced.
In the case of a data=ordered umount with pending long symlinks which
are delayed due to a long list of other I/O on the backing block
device, this causes the buffer associated with the long symlinks to
not be moved to the inode dirty list in the second phase of
fsync_super. Then, before they can be dirtied again, kjournald exits,
seeing the UMOUNT flag and the dirty pages are never written to the
backing block device, causing long symlink corruption and exposing new
or previously freed block data to userspace.
To ensure all commits are synced, we flush all journal commits now
when sync_fs'ing ext4.
Signed-off-by: Arthur Jones <ajones@riverbed.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit 23f8b79e introducd a regression because it assumed that if
there were no transactions ready to be checkpointed, that no progress
could be made on making space available in the journal, and so the
journal should be aborted. This assumption is false; it could be the
case that simply calling jbd2_cleanup_journal_tail() will recover the
necessary space, or, for small journals, the currently committing
transaction could be responsible for chewing up the required space in
the log, so we need to wait for the currently committing transaction
to finish before trying to force a checkpoint operation.
This patch fixes a bug reported by Mihai Harpau at:
https://bugzilla.redhat.com/show_bug.cgi?id=469582
This patch fixes a bug reported by François Valenduc at:
http://bugzilla.kernel.org/show_bug.cgi?id=11840
Fix a regression caused by commit d0156417, "ext4: fix ext4_dx_readdir
hash collision handling", where deleting files in a large directory
(requiring more than one getdents system call), results in some
filenames being returned twice. This was caused by a failure to
update info->curr_hash and info->curr_minor_hash, so that if the
directory had gotten modified since the last getdents() system call
(as would be the case if the user is running "rm -r" or "git clean"),
a directory entry would get returned twice to the userspace.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This patch fixes the bug reported by Markus Trippelsdorf at:
http://bugzilla.kernel.org/show_bug.cgi?id=11844
During filesystem recovery we may be doing a truncate
which expects some of the mballoc data structures to
be initialized. So do ext4_mb_init before recovery.
ext4_xattr_set_handle() eventually ends up calling
ext4_mark_inode_dirty() which tries to expand the inode by shifting
the EAs. This leads to the xattr_sem being downed again and leading
to a deadlock.
This patch makes sure that if ext4_xattr_set_handle() is in the
call-chain, ext4_mark_inode_dirty() will not expand the inode.
Also make sure the buffer heads are marked clean before submitting bh
for writing. The previous code was marking the buffer head dirty,
which would have forced an unneeded write (and seek) to the journal
for no good reason.
The __jbd2_log_wait_for_space function sits in a loop checkpointing
transactions until there is sufficient space free in the journal.
However, if there are no transactions to be processed (e.g. because the
free space calculation is wrong due to a corrupted filesystem) it will
never progress.
Check for space being required when no transactions are outstanding and
abort the journal instead of endlessly looping.
This patch fixes the bug reported by Sami Liedes at:
http://bugzilla.kernel.org/show_bug.cgi?id=10976
Pick an ioctl number for EXT4_IOC_MIGRATE that won't conflict with
other ext4 ioctl's. Since there haven't been any major userspace
users of this ioctl, we can afford to change this now, to avoid
potential problems later.
Also, reorder the ioctl numbers in ext4.h to avoid this sort of
mistake in the future.
If there group descriptors are corrupted we need unlock the block
group lock before returning from the function; else we will oops when
freeing a spinlock which is still being held.
Some devices such as "cciss/c0d0p9" will cause jbd2 setup and teardown
failures when /proc filenames are created with embedded slashes. This
is a slimmed down version of commit 05496769, with the stack reduction
aspects of the patch omitted to meet the -stable criteria.
prevent cifs_writepages() from skipping unwritten pages
Fixes a data corruption under heavy stress in which pages could be left
dirty after all open instances of a inode have been closed.
In order to write contiguous pages whenever possible, cifs_writepages()
asks pagevec_lookup_tag() for more pages than it may write at one time.
Normally, it then resets index just past the last page written before calling
pagevec_lookup_tag() again.
If cifs_writepages() can't write the first page returned, it wasn't resetting
index, and the next call to pagevec_lookup_tag() resulted in skipping all of
the pages it previously returned, even though cifs_writepages() did nothing
with them. This can result in data loss when the file descriptor is about
to be closed.
This patch ensures that index gets set back to the next returned page so
that none get skipped.
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Acked-by: Jeff Layton <jlayton@redhat.com> Cc: Shirish S Pargaonkar <shirishp@us.ibm.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Cc: Suresh Jayaraman <sjayaraman@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Use a similar approach to the SMB session sharing. Add a list of tcons
attached to each SMB session. Move the refcount to non-atomic. Protect
all of the above with the cifs_tcp_ses_lock. Add functions to
properly find and put references to the tcons.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Cc: Suresh Jayaraman <sjayaraman@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We do this by abandoning the global list of SMB sessions and instead
moving to a per-server list. This entails adding a new list head to the
TCP_Server_Info struct. The refcounting for the cifsSesInfo is moved to
a non-atomic variable. We have to protect it by a lock anyway, so there's
no benefit to making it an atomic. The list and refcount are protected
by the global cifs_tcp_ses_lock.
The patch also adds a new routines to find and put SMB sessions and
that properly take and put references under the lock.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Cc: Suresh Jayaraman <sjayaraman@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The code that allows these structs to be shared is extremely racy.
Disable the sharing of SMB and tcon structs for now until we can
come up with a way to do this that's race free.
We want to continue to share TCP sessions, however since they are
required for multiuser mounts. For that, implement a new (hopefully
race-free) scheme. Add a new global list of TCP sessions, and take
care to get a reference to it whenever we're dealing with one.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Cc: Suresh Jayaraman <sjayaraman@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We're currently declaring both a sockaddr_in and sockaddr6_in on the
stack, but we really only need storage for one of them. Declare a
sockaddr struct and cast it to the proper type. Also, eliminate the
protocolType field in the TCP_Server_Info struct. It's redundant since
we have a sa_family field in the sockaddr anyway.
We may need to revisit this if SCTP is ever implemented, but for now
this will simplify the code.
CIFS over IPv6 also has a number of problems currently. This fixes all
of them that I found. Eventually, it would be nice to move more of the
code to be protocol independent, but this is a start.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Cc: Suresh Jayaraman <sjayaraman@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Also adds two lines missing from the previous patch (for the need reconnect flag in the
/proc/fs/cifs/DebugData handling)
The new global_cifs_sock_list is added, and initialized in init_cifs but not used yet.
Jeff Layton will be adding code in to use that and to remove the GlobalTcon and GlobalSMBSession
lists.
CC: Jeff Layton <jlayton@redhat.com> CC: Shirish Pargaonkar <shirishp@us.ibm.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Cc: Suresh Jayaraman <sjayaraman@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In preparation for Jeff's big umount/mount fixes to remove the possibility of
various races in cifs mount and linked list handling of sessions, sockets and
tree connections, this patch cleans up some repetitive code in cifs_mount,
and addresses a problem with ses->status and tcon->tidStatus in which we
were overloading the "need_reconnect" state with other status in that
field. So the "need_reconnect" flag has been broken out from those
two state fields (need reconnect was not mutually exclusive from some of the
other possible tid and ses states). In addition, a few exit cases in
cifs_mount were cleaned up, and a problem with a tcon flag (for lease support)
was not being set consistently for the 2nd mount of the same share
CC: Jeff Layton <jlayton@redhat.com> CC: Shirish Pargaonkar <shirishp@us.ibm.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This is an implementation of David Miller's suggested fix in:
https://bugzilla.redhat.com/show_bug.cgi?id=470201
It has been updated to use wait_event() instead of
wait_event_interruptible().
Paraphrasing the description from the above report, it makes sendmsg()
block while UNIX garbage collection is in progress. This avoids a
situation where child processes continue to queue new FDs over a
AF_UNIX socket to a parent which is in the exit path and running
garbage collection on these FDs. This contention can result in soft
lockups and oom-killing of unrelated processes.
Signed-off-by: dann frazier <dannf@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When resizing a CQ, MTTs associated with the old CQE buffer were not
freed. As a result, if any app used resize CQ repeatedly, all MTTs
were eventually exhausted, which led to all memory registration
operations failing until the driver is reloaded.
Once the RESIZE_CQ command returns successfully from FW, FW no longer
accesses the old CQ buffer, so it is safe to deallocate the MTT
entries used by the old CQ buffer.
Finally, if the RESIZE_CQ command fails, the MTTs allocated for the
new CQEs buffer also need to be de-allocated.
This fixes <https://bugs.openfabrics.org/show_bug.cgi?id=1416>.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Add another model ID of a broken firmware to prevent early I/O errors
by acesses at the end of the disk. Reported at linux1394-user,
http://marc.info/?t=122670842900002
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Add another model ID of a broken firmware to prevent early I/O errors
by acesses at the end of the disk. Reported at linux1394-user,
http://marc.info/?t=122670842900002
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
For some unknown reason at Steven Rostedt added in disabling of the SPE
instruction generation for e500 based PPC cores in commit 6ec562328fda585be2d7f472cfac99d3b44d362a.
We are removing it because:
1. It generates e500 kernels that don't work
2. its not the correct set of flags to do this
3. we handle this in the arch/powerpc/Makefile already
4. its unknown in talking to Steven why he did this
Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Tested-and-Acked-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When the "hpwdt" module is loaded (even if the /dev/watchdog device is not
opened), then kdump does not work. The panic kernel either does not start at
all or crash in various places.
The problem is that hpwdt_pretimeout is registered with register_die_notifier()
with the highest possible priority. Because it returns NOTIFY_STOP, the
crash_nmi_callback which is also registered with register_die_notifier()
is never executed. This causes the shutdown of other CPUs to fail.
Reverting the order is no option: The crash_nmi_callback executes HLT
and so never returns normally. Because of that, it must be executed as
last notifier, which currently is done.
So, that patch returns NOTIFY_OK to keep the crash_nmi_callback executed.
Signed-off-by: Bernhard Walle <bwalle@suse.de> Signed-off-by: Wim Van Sebroeck <wim@iguana.be> Signed-off-by: Thomas Mingarelli <thomas.mingarelli@hp.com> Cc: Vivek Goyal <vgoyal@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The address provided by the SMBIOS/DMI CRU information is mapped via
ioremap() in the virtual address space. However, since the address is
executed (i.e. call'd), we need to set that pages as executable.
Without that, I get following oops on a HP ProLiant DL385 G2
machine with BIOS from 05/29/2008 when I trigger crashdump:
A mutex_unlock(&gang->aff_mutex) in spufs_create_context() is missing
in case spufs_context_open() fails. As a result, spu_create syscall
and spu_get_idle() may block.
This patch adds the mutex_unlock.
Signed-off-by: Kou Ishizaki <kou.ishizaki@toshiba.co.jp> Signed-off-by: Jeremy Kerr <jk@ozlabs.org> Acked-by: Andre Detsch <adetsch@br.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently, we can end up in an infinite loop if we get a signal
while the kernel has faulted in spufs_ps_fault. Eg:
alarm(1);
write(fd, some_spu_psmap_register_address, 4);
- the write's copy_from_user will fault on the ps mapping, and
signal_pending will be non-zero. Because returning from the fault
handler will never clear TIF_SIGPENDING, so we'll just keep faulting,
resulting in an unkillable process using 100% of CPU.
This change returns VM_FAULT_SIGBUS if there's a fatal signal pending,
letting us escape the loop.
Signed-off-by: Jeremy Kerr <jk@ozlabs.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
[CIFS] Reduce number of socket retries in large write path
CIFS in some heavy stress conditions cifs could get EAGAIN
repeatedly in smb_send2 which led to repeated retries and eventually
failure of large writes which could lead to data corruption.
There are three changes that were suggested by various network
developers:
1) convert cifs from non-blocking to blocking tcp sendmsg
(we left in the retry on failure)
2) change cifs to not set sendbuf and rcvbuf size for the socket
(let tcp autotune the buffer sizes since that works much better
in the TCP stack now)
3) if we have a partial frame sent in smb_send2, mark the tcp
session as invalid (close the socket and reconnect) so we do
not corrupt the remaining part of the SMB with the beginning
of the next SMB.
This does not appear to hurt performance measurably and has
been run in various scenarios, but it definately removes
a corruption that we were seeing in some high stress
test cases.
Acked-by: Shirish Pargaonkar <shirishp@us.ibm.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The Belkin F5D7050rev5000de (id 050d:705e) has the Realtek RTL8187B chip
and works with the 2.6.27 driver.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some recent Seagate harddrives have firmware bug which causes FLUSH
CACHE to timeout under certain circumstances if NCQ is being used.
This can be worked around by disabling NCQ and fixed by updating the
firmware. Implement ATA_HORKAGE_FIRMWARE_UPDATE and blacklist these
devices.
The wiki page has been updated to contain information on this issue.
Since dev->power.should_wakeup bit is used by the PCI core to
decide whether the device should wake up the system from sleep
states, set/unset this bit whenever WOL is enabled/disabled using
e1000_set_wol(). Accordingly, use device_can_wakeup() for checking
if wake-up is supported by the device.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since dev->power.should_wakeup bit is used by the PCI core to
decide whether the device should wake up the system from sleep
states, set/unset this bit whenever WOL is enabled/disabled using
e1000_set_wol(). Accordingly, use device_can_wakeup() for checking
if wake-up is supported by the device.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since dev->power.should_wakeup bit is used by the PCI core to
decide whether the device should wake up the system from sleep
states, set/unset this bit whenever WOL is enabled/disabled using
igb_set_wol(). Accordingly, use device_can_wakeup() for checking
if wake-up is supported by the device.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Impact: make warning message disappear - functionality unchanged
Problems with bogus IRQ0 override of those laptops should be fixed
with commits
x86: SB600: skip IRQ0 override if it is not routed to INT2 of IOAPIC
x86: SB450: skip IRQ0 override if it is not routed to INT2 of IOAPIC
that introduce early-quirks based on chipset configuration.
For further information, see
http://bugzilla.kernel.org/show_bug.cgi?id=11516
Instead of removing the related dmi-quirks completely we'd like to
keep them for (at least) one kernel version -- to double-check whether
the early-quirks really took effect. But the dmi-quirks need to be
called after early-quirks are executed. With this patch calling
sequence for dmi-quriks is changed as follows:
The NUMA code on x86_32 creates special memory mapping that allows
each node's pgdat to be located in this node's memory. For this
purpose it allocates a memory area at the end of each node's memory
and maps this area so that it is accessible with virtual addresses
belonging to low memory. As a result, if there is high memory,
these NUMA-allocated areas are physically located in high memory,
although they are mapped to low memory addresses.
Our hibernation code does not take that into account and for this
reason hibernation fails on all x86_32 systems with CONFIG_NUMA=y and
with high memory present. Fix this by adding a special mapping for
the NUMA-allocated memory areas to the temporary page tables created
during the last phase of resume.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When reserving space for the hypervisor the Xen paravirt backend adds
an extra two pages (this was carried forward from the 2.6.18-xen tree
which had them "for safety"). Depending on various CONFIG options this
can cause the boot time fixmaps to span multiple PMDs which is not
supported and triggers a WARN in early_ioremap_init().
The bad_bios_dmi_table() quirk never triggered because we do DMI setup
too late. Move it a bit earlier.
There is no real reason to reserve these two extra pages and the
fixmap already incorporates FIX_HOLE which serves the same
purpose. None of the other callers of reserve_top_address do this.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
A workaround for AMD CPU family 11h erratum 311 might cause that the
P-state Status Register shows a "current P-state" which is larger than
the "current P-state limit" in P-state Current Limit Register. For the
wrong P-state value there is no ACPI _PSS object defined and
powernow-k8/cpufreq can't determine the proper CPU frequency for that
state.
As a consequence this can cause a panic during boot (potentially with
all recent kernel versions -- at least I have reproduced it with
various 2.6.27 kernels and with the current .28 series), as an
example:
In short, aftereffect of the wrong P-state is that
cpufreq_stats_update() uses "-1" as index for some array in
cpufreq_stats_update (unsigned int cpu)
{
...
if (stat->time_in_state)
stat->time_in_state[stat->last_index] =
cputime64_add(stat->time_in_state[stat->last_index],
cputime_sub(cur_time, stat->last_time));
...
}
Fortunately, the wrong P-state value is returned only if the core is
in P-state 0. This fix solves the problem by detecting the
out-of-range P-state, ignoring it, and using "0" instead.
Cc: Mark Langsdorf <mark.langsdorf@amd.com> Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com> Signed-off-by: Dave Jones <davej@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Now that the PCI core manages the 'name' for each individual
hotplug driver, and all drivers (except rpaphp) have been converted
to use hotplug_slot_name(), there is no need for the PCI hotplug
core to drag around its own copy of name either.
We no longer need to manage our version of hotplug_slot->name
since the PCI and hotplug core manage it on our behalf.
Update the sn_hp_slot_private_alloc() interface to fill in
the correct name for us, as that function already has all
the parameters needed to determine the name.
rpaphp tends to use slot->name directly everywhere, and doesn't
ever need slot->hotplug_slot->name.
struct hotplug_slot->name is going away, so convert rpaphp directly
manipulate its own slot->name everywhere, and don't bother touching
slot->hotplug_slot->name.
In preparation for cleaning up the various hotplug drivers
such that they don't have to manage their own 'name' parameters
anymore, we provide the following convenience functions:
pci_slot_name()
hotplug_slot_name()
These helpers will be used by individual hotplug drivers.
Prevent callers of pci_create_slot() from registering slots with
duplicate names. This condition occurs most often when PCI hotplug
drivers are loaded on platforms with broken firmware that assigns
identical names to multiple slots.
We now rename these duplicate slots on behalf of the user.
If firmware assigns the name N to multiple slots, then:
The first registered slot is assigned N
The second registered slot is assigned N-1
The third registered slot is assigned N-2
etc.
This is the permanent fix mentioned in earlier commits d6a9e9b4 and 167e782e (shpchp/pciehp: Rename duplicate slot name...).
We take advantage of the new 'hotplug' parameter in pci_create_slot()
to prevent a slot create/rename race between hotplug drivers and
detection drivers.
The hotplug driver creates the slot with its desired name, and then
releases the semaphore. Now, the detection driver tries to create
the same slot, but it already exists. We don't care about renaming,
so return the existing slot.
The detection driver creates the slot with name "X". Then the hotplug
driver tries to create the same slot, but wants the name "Y" instead.
We detect that we're trying to create the same slot and that we also
want a rename, so rename the slot to "Y" and return.
Two separate hotplug drivers are attempting to claim the slot and
are passing valid hotplug_slot args to pci_create_slot(). We detect
that the slot already has a ->hotplug callback, prevent a rename,
and return -EBUSY.
Slot detection drivers can co-exist with hotplug drivers. The names
of the detected/claimed slots may be different depending on module
load order.
For legacy reasons, we need to allow hotplug drivers to override
the slot name if a detection driver is loaded first (and they find
the same slots).
Creating and overriding slot names should be an atomic operation,
otherwise you get a locking nightmare as various drivers race to
call pci_create_slot().
pci_create_slot() is already serialized by grabbing the pci_bus_sem.
We update the API and add a 'hotplug' param, which is:
set if the caller is a hotplug driver
NULL if the caller is a detection driver
pci_create_slot() does not actually use the 'hotplug' parameter in this
patch. A later patch will add the logic that uses it.
Update pci_hp_register() to take a const char *name parameter.
The motivation for this is to clean up the individual hotplug
drivers so that each one does not have to manage its own name.
The PCI core should be the place where we manage the name.
We update the interface and all callsites first, in a
"no functional change" manner, and clean up the drivers later.
after noticing that my Netgear FA411 (PCMCIA-NIC) [1] stopped working with
the release of the 2.6.25 kernel (sidux-version), I checked the
respective driver sources and noticed that the pcnet_cs driver bailed
out with "use axnet_cs instead" for the Netgear FA411, but axnet_cs
doesn't claim this ID.
I compiled a kernel with the PCMCIA-ID for the netgear card moved to
axnet_cs from pcnet_cs which worked. I then contacted sidux-kernel
maintainer Stefan Lippers-Hollmann who turned the info into this patch
and integrated it into the kernel:
This works for me and AFAIK there were no reports of any breakage for
other devices on sidux-support.
This looks like a trivial patch, but since I have very limited
experience with kernel modifications I might be woefully wrong there.
But if there are no side effects of this patch, is it possible to get it
into the official kernel?
I can provide more detailed information on the affected hardware if
necessary.
From: Stefan Lippers-Hollmann <s.l-h@gmx.de>
Date: Sat, 1 Nov 2008 23:53:04 +0000
Subject: PCMCIA: move PCMCIA ID for Netgear FA411 from pcnet_cs to axnet_cs:
Since kernel 2.6.25, commit 61da96be07ec860e260ca4af0199b9d48d000b80
(pcnet_cs: if AX88190-based card, printk "use axnet_cs instead" message.),
pcnet_cs bails out with "use axnet_cs instead" for the Netgear FA411, but
axnet_cs doesn't claim this ID.
Signed-off-by: Stefan Lippers-Hollmann <s.l-h@gmx.de> Signed-off-by: Cord Walter <qord@cwalter.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We should only tell the hardware its capable of DMA'ing
to us only what we asked dev_alloc_skb(). Prior to this
it is possible a large RX'd frame could have corrupted
DMA data but for us but we were saved only because we
were previously also pci_map_single()'ing the same large
value. The issue prior to this though was we were unmapping
a smaller amount which the prior DMA patch fixed.
Signed-off-by: Bennyam Malavazi <Bennyam.Malavazi@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This should fix the SW-IOMMU bounce buffer starvation
seen ok kernel.org bugzilla 11811:
http://bugzilla.kernel.org/show_bug.cgi?id=11811
Users on MacBook Pro 3.1/MacBook v2 would see something like:
DMA: Out of SW-IOMMU space for 4224 bytes at device 0000:0b:00.0
Unfortunately its only easy to trigger on MacBook Pro 3.1/MacBook v2
so far so its difficult to debug (even with swiotlb=force).
We were pci_unmap_single()'ing less bytes than what we called
for with pci_map_single() and as such we were starving
the swiotlb from its 64MB amount of bounce buffers. We remain
consistent and now always use sc->rxbufsize for RX. While at
it we update the beacon DMA maps as well to only use the data
portion of the skb, previous to this we were pci_map_single()'ing
more data for beaconing than what we tell the hardware it can use,
therefore pushing more iotlb abuse.
Still not sure why this is so easily triggerable on
MacBook Pro 3.1, it may be the hardware configuration
tends to use more memory > 3GB mark for DMA.
Signed-off-by: Maciej Zenczykowski <zenczykowski@gmail.com> Signed-off-by: Bennyam Malavazi <Bennyam.Malavazi@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Impact: fix boot crash on AMD IOMMU if CONFIG_GART_IOMMU is off
Currently these macros evaluate to a no-op except the kernel is compiled
with GART or Calgary support. But we also need these macros when we have
SWIOTLB, VT-d or AMD IOMMU in the kernel. Since we always compile at
least with SWIOTLB we can define these macros always.
This patch is also for stable backport for the same reason the SWIOTLB
default selection patch is.
Impact: widen the reach of the low-memory-protect DMI quirk
Phoenix BIOSes variously identify their vendor as "Phoenix Technologies,
LTD" or "Phoenix Technologies LTD" (without the comma.)
This patch makes the identification string in the bad_bios_dmi_table
more general (following a suggestion by Ingo Molnar), so that both
versions are handled.
Again, the patched file compiles cleanly and the patch has been tested
successfully on my machine.
The netmos_9xx5_combo type assumes that PCI SSID provides always the
correct value for the number of parallel and serial ports, but there are
indeed broken devices with wrong numbers, which may result in Oops.
This patch simply adds the check of the array range.