Hugh Dickins [Tue, 23 Jan 2007 15:46:22 +0000 (16:46 +0100)]
read_zero_pagealigned() locking fix
Ramiro Voicu hits the BUG_ON(!pte_none(*pte)) in zeromap_pte_range: kernel
bugzilla 7645. Right: read_zero_pagealigned uses down_read of mmap_sem,
but another thread's racing read of /dev/zero, or a normal fault, can
easily set that pte again, in between zap_page_range and zeromap_page_range
getting there. It's been wrong ever since 2.4.3.
The simple fix is to use down_write instead, but that would serialize reads
of /dev/zero more than at present: perhaps some app would be badly
affected. So instead let zeromap_page_range return the error instead of
BUG_ON, and read_zero_pagealigned break to the slower clear_user loop in
that case - there's no need to optimize for it.
Use -EEXIST for when a pte is found: BUG_ON in mmap_zero (the other user of
zeromap_page_range), though it really isn't interesting there. And since
mmap_zero wants -EAGAIN for out-of-memory, the zeromaps better return that
than -ENOMEM.
Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Alan Cox [Mon, 22 Jan 2007 19:39:00 +0000 (20:39 +0100)]
atiixp: hang fix
When the old IDE layer calls into methods in the driver during error
handling it is essentially random whether ide_lock is already held. This
causes a deadlock in the atiixp driver which also uses ide_lock internally
for locking.
Switch to a private lock instead.
[akpm@osl.org: cleanup] Signed-off-by: Alan Cox <alan@redhat.com> Acked-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Jens Axboe [Mon, 22 Jan 2007 19:34:31 +0000 (20:34 +0100)]
cdrom: set default timeout to 7 seconds
It's a known fact that Windows times out commands after 7 seconds, so
drives generally try and respond if they can before that happens. We
default to 5 seconds, which sometimes is a bit too short.
Jeremy Higdon reported here:
http://lkml.org/lkml/2007/1/1/145
that his drive takes longer than 5 seconds for a "read track
information" command, later confirming that it is about 6.7 seconds.
So just do the sane thing and change the default command timeout to 7
seconds to avoid other surprises.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Jes Sorensen [Mon, 22 Jan 2007 19:20:21 +0000 (20:20 +0100)]
[SCSI] qla1280 command timeout
Original patch from Ian Dall in bugzilla. Set command timeout as
specified by the SCSI layer rather than hardcode it to 30 seconds. I
have received a couple of reports of people hitting this one with
various tape configurations and the patch looks obviously correct.
From http://bugzilla.kernel.org/show_bug.cgi?id=6275
Ian Dall <ian@beware.dropbear.id.au>:
The command sent to the card was using a 30second timeout regardless of the
timeout requested in the scsi command passed down from higher levels.
Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
James Bursa [Sat, 20 Jan 2007 21:58:51 +0000 (22:58 +0100)]
adfs: fix filename handling
Fix filenames on adfs discs being terminated at the first character greater
than 128 (adfs filenames are Latin 1). I saw this problem when using a
loopback adfs image on a 2.6.17-rc5 x86_64 machine, and the patch fixed it
there.
Include connector config in the s390 arch Kconfig to get support for
connectors.
This also fixes the following Kconfig warning:
fs/Kconfig:1728:warning: 'select' used by config symbol 'CIFS_UPCALL' refer to undefined symbol 'CONNECTOR'
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Patrick McHardy [Sat, 20 Jan 2007 21:18:30 +0000 (22:18 +0100)]
NETFILTER: NAT: fix NOTRACK checksum handling
The whole idea with the NOTRACK netfilter target is that
you can force the netfilter code to avoid connection
tracking, and all costs assosciated with it, by making
traffic match a NOTRACK rule.
But this is totally broken by the fact that we do a checksum
calculation over the packet before we do the NOTRACK bypass
check, which is very expensive. People setup NOTRACK rules
explicitly to avoid all of these kinds of costs.
This patch from Patrick, already in Linus's tree, fixes the
bug.
Move the check for ip_conntrack_untracked before the call to
skb_checksum_help to fix NOTRACK excemptions from NAT. Pre-2.6.19
NAT code breaks TSO by invalidating hardware checksums for every
packet, even if explicitly excluded from NAT through NOTRACK.
2.6.19 includes a fix that makes NAT and TSO live in harmony,
but the performance degradation caused by this deserves making
at least the workaround work properly in -stable.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Arnaud Patard [Mon, 8 Jan 2007 22:09:59 +0000 (23:09 +0100)]
ALSA: emu10k1: Fix outl() in snd_emu10k1_resume_regs()
The emu10k1 driver saves the A_IOCFG and HCFG register on suspend and restores
it on resumes. Unfortunately, this doesn't work as the arguments to outl() are
reversed.
Jean Delvare [Mon, 8 Jan 2007 06:05:19 +0000 (07:05 +0100)]
V4L: cx88: Fix leadtek_eeprom tagging
reference to .init.text: from .text between 'cx88_card_setup'
(at offset 0x68c) and 'cx88_risc_field'
Caused by leadtek_eeprom() being declared __devinit and called from
a non-devinit context.
Signed-off-by: Jean Delvare <khali@linux-fr.org> Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Phillip Lougher [Mon, 8 Jan 2007 06:02:45 +0000 (07:02 +0100)]
corrupted cramfs filesystems cause kernel oops (CVE-2006-5823)
Steve Grubb's fzfuzzer tool (http://people.redhat.com/sgrubb/files/
fsfuzzer-0.6.tar.gz) generates corrupt Cramfs filesystems which cause
Cramfs to kernel oops in cramfs_uncompress_block(). The cause of the oops
is an unchecked corrupted block length field read by cramfs_readpage().
This patch adds a sanity check to cramfs_readpage() which checks that the
block length field is sensible. The (PAGE_CACHE_SIZE << 1) size check is
intentional, even though the uncompressed data is not going to be larger
than PAGE_CACHE_SIZE, gzip sometimes generates compressed data larger than
the original source data. Mkcramfs checks that the compressed size is
always less than or equal to PAGE_CACHE_SIZE << 1. Of course Cramfs could
use the original uncompressed data in this case, but it doesn't.
Signed-off-by: Phillip Lougher <phillip@lougher.org.uk> Signed-off-by: Adrian Bunk <bunk@stusta.de>
I've been using Steve Grubb's purely evil "fsfuzzer" tool, at
http://people.redhat.com/sgrubb/files/fsfuzzer-0.4.tar.gz
Basically it makes a filesystem, splats some random bits over it, then
tries to mount it and do some simple filesystem actions.
At best, the filesystem catches the corruption gracefully. At worst,
things spin out of control.
As you might guess, we found a couple places in ext3 where things spin out
of control :)
First, we had a corrupted directory that was never checked for
consistency... it was corrupt, and pointed to another bad "entry" of
length 0. The for() loop looped forever, since the length of
ext3_next_entry(de) was 0, and we kept looking at the same pointer over and
over and over and over... I modeled this check and subsequent action on
what is done for other directory types in ext3_readdir...
(adding this check adds some computational expense; I am testing a followup
patch to reduce the number of times we check and re-check these directory
entries, in all cases. Thanks for the idea, Andreas).
Next we had a root directory inode which had a corrupted size, claimed to
be > 200M on a 4M filesystem. There was only really 1 block in the
directory, but because the size was so large, readdir kept coming back for
more, spewing thousands of printk's along the way.
Per Andreas' suggestion, if we're in this read error condition and we're
trying to read an offset which is greater than i_blocks worth of bytes,
stop trying, and break out of the loop.
With these two changes fsfuzz test survives quite well on ext3.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Eric Sandeen [Mon, 8 Jan 2007 05:59:28 +0000 (06:59 +0100)]
ext2: skip pages past number of blocks in ext2_find_entry (CVE-2006-6054)
This one was pointed out on the MOKB site:
http://kernelfun.blogspot.com/2006/11/mokb-09-11-2006-linux-26x-ext2checkpage.html
If a directory's i_size is corrupted, ext2_find_entry() will keep processing
pages until the i_size is reached, even if there are no more blocks associated
with the directory inode. This patch puts in some minimal sanity-checking
so that we don't keep checking pages (and issuing errors) if we know there
can be no more data to read, based on the block count of the directory inode.
This is somewhat similar in approach to the ext3 patch I sent earlier this
year.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
and boom. Need to make sure the error cases return an error, I think.
[akpm@osdl.org: return -ENOMEM on oom] Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Ran into BUG() while doing madvise(REMOVE) testing. If we are punching a
hole into shared memory segment using madvise(REMOVE) and the entire hole
is below the indirect blocks, we hit following assert.
BUG_ON(limit <= SHMEM_NR_DIRECT);
Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com> Forwarded-by: Jordan Neumeyer Signed-off-by: Adrian Bunk <bunk@stusta.de>
John Heffner [Sat, 6 Jan 2007 21:31:44 +0000 (22:31 +0100)]
TCP: Fix and simplify microsecond rtt sampling
This changes the microsecond RTT sampling so that samples are taken in
the same way that RTT samples are taken for the RTO calculator: on the
last segment acknowledged, and only when the segment hasn't been
retransmitted.
Signed-off-by: John Heffner <jheffner@psc.edu> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Makes UML compile on any possible processor choice. The two problems were:
*) x86 code, when 386 is selected, checks at runtime boot_cpuflags, which we
not have.
*) 3Dnow support for memcpy() et al. does not compile currently and fixing t
is not trivial, so simply disable it; with this change, if one selects MK
UML compiles (while it did not).
Merged upstream.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Willy Tarreau [Sat, 6 Jan 2007 01:31:24 +0000 (02:31 +0100)]
rio: typo in bitwise AND expression.
The line:
hp->Mode &= !RIO_PCI_INT_ENABLE;
is obviously wrong as RIO_PCI_INT_ENABLE=0x04 and is used as a bitmask
2 lines before. Getting no IRQ would not disable RIO_PCI_INT_ENABLE
but rather RIO_PCI_BOOT_FROM_RAM which equals 0x01.
Obvious fix is to change ! for ~.
Signed-off-by: Willy Tarreau <w@1wt.eu> Signed-off-by: Adrian Bunk <bunk@stusta.de>
David Brownell [Sat, 6 Jan 2007 00:08:47 +0000 (01:08 +0100)]
SPI/MTD: mtd_dataflash oops prevention
Return a fault code if the Dataflash driver runs into a "no device present"
error when the MISO line has a pulldown (it currently expects a pullup), so
that rmmod won't oops.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Georg Chini [Sat, 6 Jan 2007 00:00:08 +0000 (01:00 +0100)]
[SOUND] Sparc CS4231: Fix IRQ return value and initialization.
SBUS: Change IRQ-handler return value from 0 to IRQ_HANDLED and
fix some initialisation problems.
Change period_bytes_min from 4096 to 256 to allow driver to work with
low latency (VOIP) applications. Hope this does not break EBUS.
Signed-off-by: Georg Chini <georg.chini@triaton-webhosting.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Andrew Morton [Thu, 4 Jan 2007 22:29:51 +0000 (23:29 +0100)]
ibmtr section fixes
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_mem_base from .text between 'ibmtr_probe1' (at offset 0x6e6) and 'ibmtr_probe_card'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_mem_base from .text between 'ibmtr_probe1' (at offset 0x74a) and 'ibmtr_probe_card'
WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_mem_base from .text between 'ibmtr_probe1' (at offset 0x7fd) and 'ibmtr_probe_card'
Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Andi Kleen [Mon, 8 Jan 2007 21:44:07 +0000 (22:44 +0100)]
x86_64: Don't leak NT bit into next task (CVE-2006-5755)
SYSENTER can cause a NT to be set which might cause crashes on the IRET
in the next task.
Following similar i386 patch from Linus.
Backport to 2.6.16 by Chuck Ebbert <76306.1226@compuserve.com>
[Changed 'set_debugreg' to the older 'set_debug' in setup64.c
and added raw_local_save_flags() from 2.6.19 to system.h]
Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Marcel Holtmann [Thu, 4 Jan 2007 21:57:52 +0000 (22:57 +0100)]
Bluetooth: Add packet size checks for CAPI messages (CVE-2006-6106)
With malformed packets it might be possible to overwrite internal
CMTP and CAPI data structures. This patch adds additional length
checks to prevent these kinds of remote attacks.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
If grow_buffers() is for some reason passed a block number which wants to li
outside the maximum-addressable pagecache range (PAGE_SIZE * 4G bytes) then
will accidentally truncate `index' and will then instnatiate a page at the
wrong pagecache offset. This causes __getblk_slow() to go into an infinite
loop.
This can happen with corrupted disks, or with software errors elsewhere.
Detect that, and handle it.
Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Linus Torvalds [Thu, 4 Jan 2007 22:23:27 +0000 (23:23 +0100)]
i386: save/restore eflags in context switch (CVE-2006-5173)
(And reset it on new thread creation)
It turns out that eflags is important to save and restore not just
because of iopl, but due to the magic bits like the NT bit, which we
don't want leaking between different threads.
Backported to 2.6.16 by Chuck Ebbert <76306.1226@compuserve.com>
[Backport consisted of removing the CFI annotations.]
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Linus Torvalds [Thu, 4 Jan 2007 00:44:45 +0000 (01:44 +0100)]
Fix incorrect user space access locking in mincore() (CVE-2006-4814)
Doug Chapman noticed that mincore() will doa "copy_to_user()" of the
result while holding the mmap semaphore for reading, which is a big
no-no. While a recursive read-lock on a semaphore in the case of a page
fault happens to work, we don't actually allow them due to deadlock
schenarios with writers due to fairness issues.
Doug and Marcel sent in a patch to fix it, but I decided to just rewrite
the mess instead - not just fixing the locking problem, but making the
code smaller and (imho) much easier to understand.
Also included are two fixes for the original patch including one
by Oleg Nesterov.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Miklos Szeredi [Thu, 4 Jan 2007 00:14:06 +0000 (01:14 +0100)]
fuse: fix hang on SMP
Fuse didn't always call i_size_write() with i_mutex held which caused
rare hangs on SMP/32bit. This bug has been present since fuse-2.2,
well before being merged into mainline.
The simplest solution is to protect i_size_write() with the
per-connection spinlock. Using i_mutex for this purpose would require
some restructuring of the code and I'm not even sure it's always safe
to acquire i_mutex in all places i_size needs to be set.
Since most of vmtruncate is already duplicated for other reasons,
duplicate the remaining part as well, making all i_size_write() calls
internal to fuse.
Using i_size_write() was unnecessary in fuse_init_inode(), since this
function is only called on a newly created locked inode.
Reported by a few people over the years, but special thanks to Dana
Henriksen who was persistent enough in helping me debug it.
Adrian Bunk:
Backported to 2.6.16.
Signed-off-by: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Dirk Eibach [Wed, 3 Jan 2007 23:42:01 +0000 (00:42 +0100)]
i2c: fix broken ds1337 initialization
On a custom board with ds1337 RTC I found that upgrade from 2.6.15 to
2.6.18 broke RTC support.
The main problem are changes to ds1337_init_client().
When a ds1337 recognizes a problem (e.g. power or clock failure) bit 7
in status register is set. This has to be reset by writing 0 to status
register. But since there are only 16 byte written to the chip and the
first byte is interpreted as an address, the status register (which is
the 16th) is never written.
The other problem is, that initializing all registers to zero is not
valid for day, date and month register. Funny enough this is checked by
ds1337_detect(), which depends on this values not being zero. So then
treated by ds1337_init_client() the ds1337 is not detected anymore,
whereas the failure bit in the status register is still set.
Patrick McHardy [Wed, 3 Jan 2007 23:38:10 +0000 (00:38 +0100)]
NET_SCHED: Fix fallout from dev->qdisc RCU change
The move of qdisc destruction to a rcu callback broke locking in the
entire qdisc layer by invalidating previously valid assumptions about
the context in which changes to the qdisc tree occur.
The two assumptions were:
- since changes only happen in process context, read_lock doesn't need
bottem half protection. Now invalid since destruction of inner qdiscs,
classifiers, actions and estimators happens in the RCU callback unless
they're manually deleted, resulting in dead-locks when read_lock in
process context is interrupted by write_lock_bh in bottem half context.
- since changes only happen under the RTNL, no additional locking is
necessary for data not used during packet processing (f.e. u32_list).
Again, since destruction now happens in the RCU callback, this assumption
is not valid anymore, causing races while using this data, which can
result in corruption or use-after-free.
Instead of "fixing" this by disabling bottem halfs everywhere and adding
new locks/refcounting, this patch makes these assumptions valid again by
moving destruction back to process context. Since only the dev->qdisc
pointer is protected by RCU, but ->enqueue and the qdisc tree are still
protected by dev->qdisc_lock, destruction of the tree can be performed
immediately and only the final free needs to happen in the rcu callback
to make sure dev_queue_xmit doesn't access already freed memory.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Trond Myklebust [Mon, 18 Dec 2006 19:55:04 +0000 (20:55 +0100)]
NFS: nfs_lookup - don't hash dentry when optimising away the lookup
If the open intents tell us that a given lookup is going to result in a,
exclusive create, we currently optimize away the lookup call itself. The
reason is that the lookup would not be atomic with the create RPC call, so
why do it in the first place?
A problem occurs, however, if the VFS aborts the exclusive create operation
after the lookup, but before the call to create the file/directory: in this
case we will end up with a hashed negative dentry in the dcache that has
never been looked up.
Fix this by only actually hashing the dentry once the create operation has
been successfully completed.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Brian King [Mon, 18 Dec 2006 19:53:26 +0000 (20:53 +0100)]
[SCSI] DAC960: PCI id table fixup
The PCI ID table in the DAC960 driver conflicts with some devices
that use the ipr driver. All ipr adapters that use this chip
have an IBM subvendor ID and all DAC960 adapters that use this
chip have a Mylex subvendor id.
Signed-off-by: Brian King <brking@us.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
bridge-netfilter: don't overwrite memory outside of skb
The bridge netfilter code needs to check for space at the
front of the skb before overwriting; otherwise if skb from
device doesn't have headroom, then it will cause random
memory corruption.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Jeff Garzik [Sun, 17 Dec 2006 23:07:43 +0000 (00:07 +0100)]
ISDN: fix drivers, by handling errors thrown by ->readstat()
This is a particularly ugly on-failure bug, possibly security, since the
lack of error handling here is covering up another class of bug: failure to
handle copy_to_user() return values.
The I4L API function ->readstat() returns an integer, and by looking at
several existing driver implementations, it is clear that a negative return
value was meant to indicate an error.
Given that several drivers already return a negative value indicating an
errno-style error, the current code would blindly accept that [negative]
value as a valid amount of bytes read. Obvious damage ensues.
Correcting ->readstat() handling to properly notice errors fixes the
existing code to work correctly on error, and enables future patches to
more easily indicate errors during operation.
Signed-off-by: Jeff Garzik <jeff@garzik.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Francois Romieu [Sun, 17 Dec 2006 21:14:09 +0000 (22:14 +0100)]
r8169: tweak the PCI data parity error recovery
The 8110SB based n2100 board signals a lot of what ought to be
PCI data parity errors durint operation of the 8169 as target.
Experiment proved that the driver can ignore the error and
process the packet as if nothing had happened.
Let's add an ad-hoc knob to enable users to fix their system while
avoiding the risks of a wholesale change.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Arjan van de Ven [Sun, 17 Dec 2006 20:49:57 +0000 (21:49 +0100)]
x86-64: Mark rdtsc as sync only for netburst, not for core2
On the Core2 cpus, the rdtsc instruction is not serializing (as defined
in the architecture reference since rdtsc exists) and due to the deep
speculation of these cores, it's possible that you can observe time go
backwards between cores due to this speculation. Since the kernel
already deals with this with the SYNC_RDTSC flag, the solution is
simple, only assume that the instruction is serializing on family 15...
The price one pays for this is a slightly slower gettimeofday (by a
dozen or two cycles), but that increase is quite small to pay for a
really-going-forward tsc counter.
Backport by Chris Wright.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Chris Wright <chrisw@sous-sol.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Robin Holt [Sun, 17 Dec 2006 20:34:56 +0000 (21:34 +0100)]
IA64: bte_unaligned_copy() transfers one extra cache line.
When called to do a transfer that has a start offset within the cache
line which is uneven between source and destination and a length which
terminates the source of the copy exactly on a cache line, one extra
line gets copied into a temporary buffer. This is normally not an issue
since the buffer is a kernel buffer and only the requested information
gets copied into the user buffer.
The problem arises when the source ends at the very last physical page
of memory. That last cache line does not exist and results in the SHUB
chip raising an MCA.
Signed-off-by: Robin Holt <holt@sgi.com> Signed-off-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Tejun Heo [Sun, 17 Dec 2006 20:32:00 +0000 (21:32 +0100)]
scsi: clear garbage after CDBs on SG_IO
ATAPI devices transfer fixed number of bytes for CDBs (12 or 16). Some
ATAPI devices choke when shorter CDB is used and the left bytes contain
garbage. Block SG_IO cleared left bytes but SCSI SG_IO didn't. This patch
makes SCSI SG_IO clear it and simplify CDB clearing in block SG_IO.
Linus Torvalds [Fri, 15 Dec 2006 00:56:30 +0000 (01:56 +0100)]
AGP: Allocate AGP pages with GFP_DMA32 by default
Not all graphic page remappers support physical addresses over the 4GB
mark for remapping, so while some do (the AMD64 GART always did, and I
just fixed the i965 to do so properly), we're safest off just forcing
GFP_DMA32 allocations to make sure graphics pages get allocated in the
low 32-bit address space by default.
AGP sub-drivers that really care, and can do better, could just choose
to implement their own allocator (or we could add another "64-bit safe"
default allocator for their use), but quite frankly, you're not likely
to care in practice.
So for now, this trivial change means that we won't be allocating pages
that we can't map correctly by mistake on x86-64.
[ On traditional 32-bit x86, this could never happen, because GFP_KERNEL
would never allocate any highmem memory anyway ]
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Neil Brown [Fri, 15 Dec 2006 00:48:58 +0000 (01:48 +0100)]
md: Fix md grow/size code to correctly find the maximum available space
An md array can be asked to change the amount of each device that it is using,
and in particular can be asked to use the maximum available space. This
currently only works if the first device is not larger than the rest. As
'size' gets changed and so 'fit' becomes wrong. So check if a 'fit' is
required early and don't corrupt it.
Signed-off-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Zachary Amsden [Fri, 15 Dec 2006 00:38:04 +0000 (01:38 +0100)]
softirq: remove BUG_ONs which can incorrectly trigger
It is possible to have tasklets get scheduled before softirqd has had a chance
to spawn on all CPUs. This is totally harmless; after success during action
CPU_UP_PREPARE, action CPU_ONLINE will be called, which immediately wakes
softirqd on the appropriate CPU to process the already pending tasklets. So
there is no danger of having a missed wakeup for any tasklets that were
already pending.
In particular, i386 is affected by this during startup, and is visible when
using a very large initrd; during the time it takes for the initrd to be
decompressed, a timer IRQ can come in and schedule RCU callbacks. It is also
possible that resending of a hardware IRQ via a softirq triggers the same bug.
Because of different timing conditions, this shows up in all emulators and
virtual machines tested, including Xen, VMware, Virtual PC, and Qemu. It is
also possible to trigger on native hardware with a large enough initrd,
although I don't have a reliable case demonstrating that.
Signed-off-by: Zachary Amsden <zach@vmware.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Christophe Saout [Fri, 15 Dec 2006 00:21:59 +0000 (01:21 +0100)]
dm crypt: Fix data corruption with dm-crypt over RAID5
Fix corruption issue with dm-crypt on top of software raid5. Cancelled
readahead bio's that report no error, just have BIO_UPTODATE cleared
were reported as successful reads to the higher layers (and leaving
random content in the buffer cache). Already fixed in 2.6.19.
Signed-off-by: Christophe Saout <christophe@saout.de> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Christophe Saout [Fri, 15 Dec 2006 00:20:35 +0000 (01:20 +0100)]
Fix SUNRPC wakeup/execute race condition
The sunrpc scheduler contains a race condition that can let an RPC
task end up being neither running nor on any wait queue. The race takes
place between rpc_make_runnable (called from rpc_wake_up_task) and
__rpc_execute under the following condition:
First __rpc_execute calls tk_action which puts the task on some wait
queue. The task is dequeued by another process before __rpc_execute
continues its execution. While executing rpc_make_runnable exactly after
setting the task `running' bit and before clearing the `queued' bit
__rpc_execute picks up execution, clears `running' and subsequently
both functions fall through, both under the false assumption somebody
else took the job.
Swapping rpc_test_and_set_running with rpc_clear_queued in
rpc_make_runnable fixes that hole. This introduces another possible
race condition that can be handled by checking for `queued' after
setting the `running' bit.
Bug noticed on a 4-way x86_64 system under XEN with an NFSv4 server
on the same physical machine, apparently one of the few ways to hit
this race condition at all.
Mark McLoughlin [Thu, 14 Dec 2006 22:09:07 +0000 (23:09 +0100)]
dm snapshot: fix metadata writing when suspending
When suspending a device-mapper device, dm_suspend() sleeps until all
necessary I/O is completed. This state is triggered by a callback from
persistent_commit(). But some I/O can still be issued *after* the callback
(to prepare the next metadata area for use if the current one is full). This
patch delays the callback until after that I/O is complete.
Signed-off-by: Mark McLoughlin <markmc@redhat.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Daniel Kobras [Thu, 14 Dec 2006 22:08:32 +0000 (23:08 +0100)]
dm: Fix deadlock under high i/o load in raid1 setup.
On an nForce4-equipped machine with two SATA disk in raid1 setup using dmraid,
we experienced frequent deadlock of the system under high i/o load. 'cat
/dev/zero > ~/zero' was the most reliable way to reproduce them: Randomly
after a few GB, 'cp' would be left in 'D' state along with kjournald and
kmirrord. The functions cp and kjournald were blocked in did vary, but
kmirrord's wchan always pointed to 'mempool_alloc()'. We've seen this pattern
on 2.6.15 and 2.6.17 kernels. http://lkml.org/lkml/2005/4/20/142 indicates
that this problem has been around even before.
So much for the facts, here's my interpretation: mempool_alloc() first tries
to atomically allocate the requested memory, or falls back to hand out
preallocated chunks from the mempool. If both fail, it puts the calling
process (kmirrord in this case) on a private waitqueue until somebody refills
the pool. Where the only 'somebody' is kmirrord itself, so we have a
deadlock.
I worked around this problem by falling back to a (blocking) kmalloc when
before kmirrord would have ended up on the waitqueue. This defeats part of
the benefits of using the mempool, but at least keeps the system running. And
it could be done with a two-line change. Note that mempool_alloc() clears the
GFP_NOIO flag internally, and only uses it to decide whether to wait or return
an error if immediate allocation fails, so the attached patch doesn't change
behaviour in the non-deadlocking case. Path is against current git
(2.6.18-rc4), but should apply to earlier versions as well. I've tested on
2.6.15, where this patch makes the difference between random lockup and a
stable system.
Signed-off-by: Daniel Kobras <kobras@linux.de> Acked-by: Alasdair G Kergon <agk@redhat.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Neil Brown [Thu, 14 Dec 2006 22:07:54 +0000 (23:07 +0100)]
dm: mirror sector offset fix
The device-mapper core does not perform any remapping of bios before passing
them to the targets. If a particular mapping begins part-way into a device,
targets obtain the sector relative to the start of the mapping by subtracting
ti->begin.
The dm-raid1 target didn't do this everywhere: this patch fixes it, taking
care to subtract ti->begin exactly once for each bio.
Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Jeff Mahoney [Thu, 14 Dec 2006 22:07:12 +0000 (23:07 +0100)]
dm: add module ref counting
The reference counting on dm-mod is zero if no mapped devices are open. This
is incorrect, and can lead to an oops if the module is unloaded while mapped
devices exist.
This patch claims a reference to the module whenever a device is created, and
drops it again when the device is freed.
Devices must be removed before dm-mod is unloaded.
Signed-off-by: Jeff Mahoney <jeffm@suse.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Persistent snapshots currently store a private copy of the chunk size.
Userspace also supplies the chunk size when loading a snapshot. Ensure
consistency by only storing the chunk_size in one place instead of two.
Currently the two sizes will differ if the chunk size supplied by userspace
does not match the chunk size an existing snapshot actually uses. Amongst
other problems, this causes an incorrect 'percentage full' to be reported.
The patch ensures consistency by only storing the chunk_size in one place,
removing it from struct pstore. Some initialisation is delayed until the
correct chunk_size is known. If read_header() discovers that the wrong chun
size was supplied, the 'area' buffer (which the header already got read into
is reinitialised to the correct size.
Signed-off-by: Alasdair G Kergon <agk@redhat.com> Signed-off-by: Adrian Bunk <bunk@stusta.de>
Michal Miroslaw [Thu, 14 Dec 2006 22:02:33 +0000 (23:02 +0100)]
dm: BUG/OOPS fix
Fix BUG I tripped on while testing failover and multipathing.
BUG shows up on error path in multipath_ctr() when parse_priority_group()
fails after returning at least once without error. The fix is to
initialize m->ti early - just after alloc()ing it.
projects / karo-tx-linux.git / log