Ever since commit 4c563f7669c10a12354b72b518c2287ffc6ebfb3
("[XFRM]: Speed up xfrm_policy and xfrm_state walking") it is
illegal to call __xfrm_state_destroy (and thus xfrm_state_put())
with xfrm_state_lock held. If we do, we'll deadlock since we
have the lock already and __xfrm_state_destroy() tries to take
it again.
Fix this by pushing the xfrm_state_put() calls after the lock
is dropped.
Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The number of identifiers needs to be checked against the option
length. Also, the identifier index provided needs to be verified
to make sure that it doesn't exceed the bounds of the array.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The bonds check to prevent buffer overlflow was not exactly
right. It still allowed overflow of up to 8 bytes which is
sizeof(struct sctp_authkey).
Since optlen is already checked against the size of that struct,
we are guaranteed not to cause interger overflow either.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
which broke userpsace, in particular the XORP build as reported by
Jose Calhariz, the debain package maintainer for XORP.
Nothing originally in linux/mroute.h was exported to userspace
ever, but some of this stuff started to be when it was moved into
this new linux/pim.h, and that was wrong. If we didn't provide these
definitions for 10 years we can reasonable expect that applications
defined this stuff locally or used GLIBC headers providing the
protocol definitions. And as such the only result of this can
be conflict and userland build breakage.
The commit #1 had such a short and terse commit message, that we
cannot even know why such a move and set of new userland exports were
even made.
Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
[ No upstream commit, this is fixing code no longer in 2.6.27 ]
nla_parse_nested_compat() was used to parse two different message
formats in the netem and prio qdisc, when it was "fixed" to work
with netem, it broke the multi queue support in the prio qdisc.
Since the prio qdisc code in question is already removed in the
development tree, this patch only fixes the regression in the
stable tree.
Based on original patch from Alexander H Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Thomas Graf <tgraf@suug.ch> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The structure used for SCTP_AUTH_KEY option contains a
length that needs to be verfied to prevent buffer overflow
conditions. Spoted by Eugene Teo <eteo@redhat.com>.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Eugene Teo <eugeneteo@kernel.sg> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
All of the SCTP-AUTH socket options could cause a panic
if the extension is disabled and the API is envoked.
Additionally, there were some additional assumptions that
certain pointers would always be valid which may not
always be the case.
This patch hardens the API and address all of the crash
scenarios.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The socket lock is there to protect the normal UDP receive path.
Encapsulation UDP sockets don't need that protection. In fact
the locking is deadly for them as they may contain another UDP
packet within, possibly with the same addresses.
Also the nested bit was copied from TCP. TCP needs it because
of accept(2) spawning sockets. This simply doesn't apply to UDP
so I've removed it.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When an action is added several times with the same exact index
it gets deleted on every even-numbered attempt.
This fixes that issue.
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Packet schedulers should only return NET_XMIT_DROP iff
the packet really was dropped. If the packet does reach
the device after we return NET_XMIT_DROP then TCP can
crash because it depends upon the enqueue path return
values being accurate.
Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ipv6_dev_get_saddr() blindly de-references dst_dev to get the network
namespace, but some callers might pass NULL. Change callers to pass a
namespace pointer instead.
Signed-off-by: Brian Haley <brian.haley@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alexey Dobriyan wrote:
> On Thu, Aug 07, 2008 at 07:00:56PM +0200, John Gumb wrote:
>> Scenario: no ipv6 default route set.
>
>> # ip -f inet6 route get fec0::1
>>
>> BUG: unable to handle kernel NULL pointer dereference at 00000000
>> IP: [<c0369b85>] rt6_fill_node+0x175/0x3b0
>> EIP is at rt6_fill_node+0x175/0x3b0
>
> 0xffffffff80424dd3 is in rt6_fill_node (net/ipv6/route.c:2191).
> 2186 } else
> 2187 #endif
> 2188 NLA_PUT_U32(skb, RTA_IIF, iif);
> 2189 } else if (dst) {
> 2190 struct in6_addr saddr_buf;
> 2191 ====> if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
> ^^^^^^^^^^^^^^^^^^^^^^^^
> NULL
>
> 2192 dst, 0, &saddr_buf) == 0)
> 2193 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
> 2194 }
The commit that changed this can't be reverted easily, but the patch
below works for me.
Fix NULL de-reference in rt6_fill_node() when there's no IPv6 input
device present in the dst entry.
Signed-off-by: Brian Haley <brian.haley@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since 49ffcf8f99e8d33ec8afb450956804af518fd788 ("sysctl: update
sysctl_check_table") setting struct ctl_table.procname = NULL does no
longer work as it used to the way the AX.25 code is expecting it to
resulting in the AX.25 sysctl registration code to break if
CONFIG_AX25_DAMA_SLAVE was not set as in some distribution kernels.
Kernel releases from 2.6.24 are affected.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I have gotten to the root cause of the hugetlb badness I reported back on
August 15th. My system has the following memory topology (note the
overlapping node):
setup_zone_migrate_reserve() scans the address range 0x0-0x8000000 looking
for a pageblock to move onto the MIGRATE_RESERVE list. Finding no
candidates, it happily continues the scan into 0x8000000-0x44000000. When
a pageblock is found, the pages are moved to the MIGRATE_RESERVE list on
the wrong zone. Oops.
setup_zone_migrate_reserve() should skip pageblocks in overlapping nodes.
Signed-off-by: Adam Litke <agl@us.ibm.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Cc: Dave Hansen <dave@linux.vnet.ibm.com> Cc: Nishanth Aravamudan <nacc@us.ibm.com> Cc: Andy Whitcroft <apw@shadowen.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>
Recent changes to tighten the check for UARTs that don't correctly
re-assert THRE (01c194d9278efc15d4785ff205643e9c0bdcef53: "serial 8250:
tighten test for using backup timer") caused problems when such a UART was
opened for the second time - the bug could only successfully be detected
at first initialization. For users of this version of this particular
UART IP it is fatal.
This patch stores the information about the bug in the bugs field of the
port structure when the port is first started up so subsequent opens can
check this bit even if the test for the bug fails.
David Brownell: "My own exposure to this is that the UART on DaVinci
hardware, which TI allegedly derived from its original 16550 logic, has
periodically gone from working to unusable with the mainline 8250.c ...
and back and forth a bunch. Currently it's "unusable", a regression from
some previous versions. With this patch from Will, it's usable."
Signed-off-by: Will Newton <will.newton@gmail.com> Acked-by: Alex Williamson <alex.williamson@hp.com> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: David Brownell <david-b@pacbell.net> 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>
spin_lock_irqsave(&r->lock, flags);
...
if (r->entropy_count > r->poolinfo->POOLBITS)
r->entropy_count = r->poolinfo->POOLBITS;
so there is a time window in which this BUG_ON():
static size_t account(struct entropy_store *r, size_t nbytes, int min,
int reserved)
{
unsigned long flags;
BUG_ON(r->entropy_count > r->poolinfo->POOLBITS);
/* Hold lock while accounting */
spin_lock_irqsave(&r->lock, flags);
can trigger.
We could fix this by moving the assertion inside the lock, but it seems
safer and saner to revert to the old behaviour wherein
entropy_store.entropy_count at no time exceeds
entropy_store.poolinfo->POOLBITS.
The direct I/O write codepath for CIFS is done through
cifs_user_write(). That function does not currently call
generic_write_checks() so the file position isn't being properly set
when the file is opened with O_APPEND. It's also not doing the other
"normal" checks that should be done for a write call.
The problem is currently that when you open a file with O_APPEND on a
mount with the directio mount option, the file position is set to the
beginning of the file. This makes any subsequent writes clobber the data
in the file starting at the beginning.
This seems to fix the problem in cursory testing. It is, however
important to note that NFS disallows the combination of
(O_DIRECT|O_APPEND). If my understanding is correct, the concern is
races with multiple clients appending to a file clobbering each others'
data. Since the write model for CIFS and NFS is pretty similar in this
regard, CIFS is probably subject to the same sort of races. What's
unclear to me is why this is a particular problem with O_DIRECT and not
with buffered writes...
Regardless, disallowing O_APPEND on an entire mount is probably not
reasonable, so we'll probably just have to deal with it and reevaluate
this flag combination when we get proper support for O_DIRECT. In the
meantime this patch at least fixes the existing problem.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix the checksum feature advertised in device flags. The hardware support
TCP/UDP over IPv4 and TCP/UDP over IPv6 (without IPv6 extension headers).
However, the kernel feature flags do not distinguish IPv6 with/without
extension headers.
Therefore, the driver needs to use NETIF_F_IP_CSUM instead of
NETIF_F_HW_CSUM since the latter includes all IPv6 packets.
A future patch can be created to check for extension headers and perform
software checksum calculation.
Signed-off-by: Ayaz Abdulla <aabdulla@nvidia.com> Cc: Jeff Garzik <jgarzik@pobox.com> Cc: Manfred Spraul <manfred@colorfullife.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
bio_copy_kern() uses bio->bv_len to copy data for READ commands after
the completion but it doesn't work with a request that partially
completed. SCSI always completes a PC request as a whole but seems
some don't.
This patch fixes bio_copy_kern to handle the above case. As
bio_copy_user does, bio_copy_kern uses struct bio_map_data to store
struct bio_vec.
__bio_copy_iov() uses bio->bv_len to copy data for READ commands after
the completion but it doesn't work with a request that partially
completed. SCSI always completes a PC request as a whole but seems
some don't.
The Perform Subsystem Function/Prepare for Read Subsystem Data
command requires 12 bytes of parameter data, but the respective data
structure dasd_psf_prssd_data has a length of 16 bytes.
Current storage servers ignore the obsolete bytes, but older models
fail to execute the command and report an incorrect length error.
This causes the device initilization for these devices to fail.
To fix this problem we need to correct the dasd_psf_prssd_data
structure and shorten it to the correct length.
Reported-by: Ivan Warren <ivan@vmfacility.fr> Reviewed-by: Ivan Warren <ivan@vmfacility.fr> Tested-by: Ivan Warren <ivan@vmfacility.fr> Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Cyrix MediaGXm/Cx5530 Unicorn Revision 1.19.3B has stopped
booting starting at v2.6.22.
The reason is this commit:
> commit f25f64ed5bd3c2932493681bdfdb483ea707da0a
> Author: Juergen Beisert <juergen@kreuzholzen.de>
> Date: Sun Jul 22 11:12:38 2007 +0200
>
> x86: Replace NSC/Cyrix specific chipset access macros by inlined functions.
this commit activated a macro which was dormant before due to (buggy)
macro side-effects.
I've looked through various datasheets and found that the GXm and GXLV
Geode processors don't have an incrementor.
Remove the incrementor setup entirely. As the incrementor value
differs according to clock speed and we would hope that the BIOS
configures it correctly, it is probably the right solution.
Joshua Hoblitt reported that only 3 GB of his 16 GB of RAM is
usable. Booting with mtrr_show showed us the BIOS-initialized
MTRR settings - which are all wrong.
So the root cause is that the BIOS has not set the mask correctly:
On Wed, Jul 23, 2008 at 03:52:36PM +0300, Andrei Popa wrote:
> I installed gnokii-0.6.22-r2 and gave the command "gnokii --identify"
> and the kernel oopsed:
>
> BUG: unable to handle kernel NULL pointer dereference at 00000458
> IP: [<c0444b52>] mutex_unlock+0x0/0xb
> [<c03830ae>] acm_tty_open+0x4c/0x214
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Tested-by: Andrei Popa <andrei.popa@i-neo.ro> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Date: Wed, 20 Aug 2008 14:09:10 -0700
Subject: binfmt_misc: fix false -ENOEXEC when coupled with other binary handlers
In case the binfmt_misc binary handler is registered *before* the e.g.
script one (when for example being compiled as a module) the following
situation may occur:
1. user launches a script, whose interpreter is a misc binary;
2. the load_misc_binary sets the misc_bang and returns -ENOEVEC,
since the binary is a script;
3. the load_script_binary loads one and calls for search_binary_hander
to run the interpreter;
4. the load_misc_binary is called again, but refuses to load the
binary due to misc_bang bit set.
The fix is to move the misc_bang setting lower - prior to the actual
call to the search_binary_handler.
Caused by the commit 3a2e7f47 (binfmt_misc.c: avoid potential kernel
stack overflow)
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Reported-by: Kirill A. Shutemov <kirill@shutemov.name> Tested-by: Kirill A. Shutemov <kirill@shutemov.name> 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>
Alex Chiang and Matthew Wilcox pointed out that pci_get_dev_by_id() does
not properly decrement the reference on the from pointer if it is
present, like the documentation for the function states it will.
It fixes a pretty bad leak in the hotplug core (we were leaking an
entire struct pci_dev for each function of each offlined card, the first
time around; subsequent onlines/offlines were ok).
Cc: Matthew Wilcox <matthew@wil.cx> Tested-by: Alex Chiang <achiang@hp.com> Acked-by: Alex Chiang <achiang@hp.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
After commit a97c9bf33f4612e2aed6f000f6b1d268b6814f3c (fix cramfs
making duplicate entries in inode cache) in kernel 2.6.14, named-pipe
on cramfs does not work properly.
It seems the commit make all named-pipe on cramfs share their inode
(and named-pipe buffer).
Make ..._test() refuse to merge inodes with ->i_ino == 1, take inode setup
back to get_cramfs_inode() and make ->drop_inode() evict ones with ->i_ino
== 1 immediately.
Reported-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp> Cc: Al Viro <viro@zeniv.linux.org.uk> 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>
Wolfgang also found out that adding kernel_fpu_begin() and kernel_fpu_end()
around the padlock instructions fix the oops.
Suresh wrote:
These padlock instructions though don't use/touch SSE registers, but it behaves
similar to other SSE instructions. For example, it might cause DNA faults
when cr0.ts is set. While this is a spurious DNA trap, it might cause
oops with the recent fpu code changes.
This is the code sequence that is probably causing this problem:
a) new app is getting exec'd and it is somewhere in between
start_thread() and flush_old_exec() in the load_xyz_binary()
b) At pont "a", task's fpu state (like TS_USEDFPU, used_math() etc) is
cleared.
c) Now we get an interrupt/softirq which starts using these encrypt/decrypt
routines in the network stack. This generates a math fault (as
cr0.ts is '1') which sets TS_USEDFPU and restores the math that is
in the task's xstate.
d) Return to exec code path, which does start_thread() which does
free_thread_xstate() and sets xstate pointer to NULL while
the TS_USEDFPU is still set.
e) At the next context switch from the new exec'd task to another task,
we have a scenarios where TS_USEDFPU is set but xstate pointer is null.
This can cause an oops during unlazy_fpu() in __switch_to()
Now:
1) This should happen with or with out pre-emption. Viro also encountered
similar problem with out CONFIG_PREEMPT.
2) kernel_fpu_begin() and kernel_fpu_end() will fix this problem, because
kernel_fpu_begin() will manually do a clts() and won't run in to the
situation of setting TS_USEDFPU in step "c" above.
3) This was working before the fpu changes, because its a spurious
math fault which doesn't corrupt any fpu/sse registers and the task's
math state was always in an allocated state.
With out the recent lazy fpu allocation changes, while we don't see oops,
there is a possible race still present in older kernels(for example,
while kernel is using kernel_fpu_begin() in some optimized clear/copy
page and an interrupt/softirq happens which uses these padlock
instructions generating DNA fault).
This is the failing scenario that existed even before the lazy fpu allocation
changes:
0. CPU's TS flag is set
1. kernel using FPU in some optimized copy routine and while doing
kernel_fpu_begin() takes an interrupt just before doing clts()
2. Takes an interrupt and ipsec uses padlock instruction. And we
take a DNA fault as TS flag is still set.
3. We handle the DNA fault and set TS_USEDFPU and clear cr0.ts
4. We complete the padlock routine
5. Go back to step-1, which resumes clts() in kernel_fpu_begin(), finishes
the optimized copy routine and does kernel_fpu_end(). At this point,
we have cr0.ts again set to '1' but the task's TS_USEFPU is stilll
set and not cleared.
6. Now kernel resumes its user operation. And at the next context
switch, kernel sees it has do a FP save as TS_USEDFPU is still set
and then will do a unlazy_fpu() in __switch_to(). unlazy_fpu()
will take a DNA fault, as cr0.ts is '1' and now, because we are
in __switch_to(), math_state_restore() will get confused and will
restore the next task's FP state and will save it in prev tasks's FP state.
Remember, in __switch_to() we are already on the stack of the next task
but take a DNA fault for the prev task.
This causes the fpu leakage.
Fix the padlock instruction usage by calling them inside the
context of new routines irq_ts_save/restore(), which clear/restore cr0.ts
manually in the interrupt context. This will not generate spurious DNA
in the context of the interrupt which will fix the oops encountered and
the possible FPU leakage issue.
Reported-and-bisected-by: Wolfgang Walter <wolfgang.walter@stwm.de> Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
[CIFS] properly account for new user= field in SPNEGO upcall string allocation
...it doesn't look like it's being accounted for at the moment. Also
try to reorganize the calculation to make it a little more evident
what each piece means.
This should probably go to the stable series as well...
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1119b) will help to reduce the clutter of usb-storage's
unusual_devs file by automatically detecting some devices that need
the IGNORE_RESIDUE flag. The idea is that devices should never return
a non-zero residue for an INQUIRY or a READ CAPACITY command unless
they failed to transfer all the requested data. So if one of these
commands transfers a standard amount of data but there is a positive
residue, we know that the residue is bogus and we can set the flag.
This fixes the problems reported in Bugzilla #11125.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Tested-by: Matthew Frost <artusemrys@sbcglobal.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1110) reverts an earlier patch meant to help with
Wireless USB host controllers. These controllers can have bulk
maxpacket values larger than 512, which puts unusual constraints on
the sizes of scatter-gather list elements. However it turns out that
the block layer does not provide the support we need to enforce these
constraints; merely changing the DMA alignment mask doesn't help.
Hence there's no reason to keep the original patch. The Wireless USB
problem will have to be solved a different way.
In addition, there is a reason to get rid of the earlier patch. By
dereferencing a pointer stored in the ep_in array of struct
usb_device, the current code risks an invalid memory access when it
runs concurrently with device removal. The members of that array are
cleared before the driver's disconnect method is called, so it should
not try to use them.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As there's no point in adding a fixed-fudge value (originally 5
seconds), honor the user settings only. We also remove the
driver's dead-callback get_rport_dev_loss_tmo function
(qla2x00_get_rport_loss_tmo()).
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
yesterday I tried to reactivate my old 486 box and wanted to install a
current Linux with latest kernel on it. But it turned out that the
latest kernel does not boot because the machine crashes early in the
setup code.
After some debugging it turned out that the problem is the query_ist()
function. If this interrupt with that function is called the machine
simply locks up. It looks like a BIOS bug. Looking for a workaround for
this problem I wrote the attached patch. It checks for the CPUID
instruction and if it is not implemented it does not call the speedstep
BIOS function. As far as I know speedstep should be available since some
Pentium earliest.
Alan Cox observed that it's available since the Pentium II, so cpuid
levels 4 and 5 can be excluded altogether.
H. Peter Anvin cleaned up the code some more:
> Right in concept, but I dislike the implementation (duplication of the
> CPU detect code we already have). Could you try this patch and see if
> it works for you?
which, with a small modification to fix a build error with it the
resulting kernel boots on my machine.
Record one more level of stack frame program counter.
Particularly when lockdep and all sorts of spinlock debugging is
enabled, figuring out the caller of spin_lock() is difficult when the
cpu is stuck on the lock.
Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The calls down into prom_printf() when we detect an overflowed stack
can recurse again since the overflow stack will be "below" the current
kernel stack limit.
Prevent this by just returning straight if we are on the stack
overflow safe stack already.
Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Larry Finger [Wed, 6 Aug 2008 04:20:56 +0000 (23:20 -0500)]
rtl8187: Fix lockups due to concurrent access to config routine
With the rtl8187 driver, the config routine is not protected against
access before a previous call has completed. When this happens, the
TX loopback that is needed to change channels may cause the chip to
be locked with a reset needed to restore communications. This patch
entered mainline as commit 7dcdd073bf78bb6958bbc12a1a47754a0f3c4721.
The problem was found by Herton Ronaldo Krzesinski <herton@mandriva.com.br>,
who also suggested this type of fix.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Acked-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br> Acked-by: Hin-Tak Leung <htl10@users.sourceforge.net> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I have a new PCI-E radeon RV380 series card (PCI device ID 5b64) that
hangs in my sparc64 boxes when the init scripts set the font. The problem
goes away if I disable acceleration.
I haven't figured out that bug yet, but along the way I found some
corrections to make based upon some auditing.
1) The RB2D_DC_FLUSH_ALL value used by the kernel fb driver
and the XORG video driver differ. I've made the kernel
match what XORG is using.
2) In radeonfb_engine_reset() we have top-level code structure
that roughly looks like:
if (family is 300, 350, or V350)
do this;
else
do that;
...
if (family is NOT 300, OR
family is NOT 350, OR
family is NOT V350)
do another thing;
this last conditional makes no sense, is always true,
and obviously was likely meant to be "family is NOT
300, 350, or V350". So I've made the code match the
intent.
Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Tested-by: Benjamin Herrenschmidt <benh@kernel.crashing.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>
The driver of ITE8212 in pass-through mode (it8212.noraid=1 on cmndline)
attempts to use the field `.dma_host_set' of the struct ide_dma_ops in
`ide_config_drive_speed' which is set to NULL by default.
So give a value to all fields of the struct ide_dma_ops.
If an OSS application calls SNDCTL_DSP_SYNC, then ALSA will call the driver's
_hw_params and _prepare functions again. On the Freescale MPC8610 DMA ASoC
driver, this caused the DMA controller to be unneccessarily re-programmed, and
apparently it doesn't like that. The DMA will then not operate when
instructed. This patch relocates much of the DMA programming to
fsl_dma_open(), which is called only once.
Signed-off-by: Timur Tabi <timur@freescale.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In kernel 2.6.26, the ability to select I2C algorithm drivers manually
was removed, as all in-kernel drivers do that automatically. However
there were some complaints that it was a problem for out-of-tree I2C
bus drivers. In order to address these complaints, let's allow manual
selection of these drivers again, but still hide them by default for
better general user experience.
This closes bug #11140:
http://bugzilla.kernel.org/show_bug.cgi?id=11140
Signed-off-by: Jean Delvare <khali@linux-fr.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix a NULL pointer dereference that happened when calling
i2c_new_probed_device on one of the addresses for which we use byte
reads instead of quick write for detection purpose (that is: 0x30-0x37
and 0x50-0x5f).
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Jean Delvare <khali@linux-fr.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The Freescale MPC8610 SSI device has the option of using one clock for both
transmit and receive (synchronous mode), or independent clocks (asynchronous).
The SSI driver, however, programs the SSI into synchronous mode and then
tries to program the clock registers independently. The result is that the wrong
sample size is usually generated during recording.
This patch fixes the discrepancy by restricting the sample rate and sample size
of the playback and capture streams. The SSI driver remembers which stream
is opened first. When a second stream is opened, that stream is constrained
to the same sample rate and size as the first stream.
A future version of this driver will lift the sample size restriction.
Supporting independent sample rates is more difficult, because only certain
codecs provide dual independent clocks.
Signed-off-by: Timur Tabi <timur@freescale.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The magic write to register 0x82 will often cause PCI config space on
my 8168 (PCI ID 10ec:8168, revision 2. mounted in an LG P300 laptop)
to be filled with ones during driver load, and thus breaking NIC
operation until reboot. If it does not happen on first driver load it
can easily be reproduced by unloading and loading the driver a few
times.
The magic write was added long ago by this commit:
Author: François Romieu <romieu@fr.zoreil.com>
Date: Sat Jan 10 06:00:46 2004 -0500
[netdrvr r8169] Merge of changes done by Realtek to rtl8169_init_one():
- phy capability settings allows lower or equal capability as suggested
in Realtek's changes;
- I/O voodoo;
- no need to s/mdio_write/RTL8169_WRITE_GMII_REG/;
- s/rtl8169_hw_PHY_config/rtl8169_hw_phy_config/;
- rtl8169_hw_phy_config(): ad-hoc struct "phy_magic" to limit duplication
of code (yep, the u16 -> int conversions should work as expected);
- variable renames and whitepace changes ignored.
As the 8168 wasn't supported by that version this patch simply removes
the bogus write from mac versions <= RTL_GIGA_MAC_VER_06.
[The change above makes sense for the 8101/8102 too -- Ueimor]
There is a call to local_irq_restore in the normal exit case, so it would
seem that there should be one on an error return as well.
The semantic patch that finds this problem is as follows:
(http://www.emn.fr/x-info/coccinelle/)
// <smpl>
@@
expression l;
expression E,E1,E2;
@@
local_irq_save(l);
... when != local_irq_restore(l)
when != spin_unlock_irqrestore(E,l)
when any
when strict
(
if (...) { ... when != local_irq_restore(l)
when != spin_unlock_irqrestore(E1,l)
+ local_irq_restore(l);
return ...;
}
|
if (...)
+ {local_irq_restore(l);
return ...;
+ }
|
spin_unlock_irqrestore(E2,l);
|
local_irq_restore(l);
)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk> Signed-off-by: Avi Kivity <avi@qumranet.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When an event (such as an interrupt) is injected, and the stack is
shadowed (and therefore write protected), the guest will exit. The
current code will see that the stack is shadowed and emulate a few
instructions, each time postponing the injection. Eventually the
injection may succeed, but at that time the guest may be unwilling
to accept the interrupt (for example, the TPR may have changed).
This occurs every once in a while during a Windows 2008 boot.
Fix by unshadowing the fault address if the fault was due to an event
injection.
Signed-off-by: Avi Kivity <avi@qumranet.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is no guarantee that the old TSS descriptor in the GDT contains
the proper base address. This is the case for Windows installation's
reboot-via-triplefault.
Use guest registers instead. Also translate the address properly.
Correct sparc64's implementation of FUTEX_OP_ANDN to do a
bitwise negate of the oparg parameter before applying the
AND operation. All other archs that support FUTEX_OP_ANDN
either negate oparg explicitly (frv, ia64, mips, sh, x86),
or do so indirectly by using an and-not instruction (powerpc).
Since sparc64 has and-not, I chose to use that solution.
I've not found any use of FUTEX_OP_ANDN in glibc so the
impact of this bug is probably minor. But other user-space
components may try to use it so it should still get fixed.
Signed-off-by: Mikael Pettersson <mikpe@it.uu.se> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
SCTP used ip6_xmit() to send fragments after received ICMP packet too
big message. But while send packet used ip6_xmit, the skb->local_df is
not initialized. So when skb if enter ip6_fragment(), the following
code will discard the skb.
SCTP do the following step:
1. send packet ip6_xmit(skb, ipfragok=0)
2. received ICMP packet too big message
3. if PMTUD_ENABLE: ip6_xmit(skb, ipfragok=1)
This patch fixed the problem by set local_df if ipfragok is true.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com> Acked-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The rationale is:
* use u32 consistently
* no need to do LCG on values from (better) get_random_bytes
* use more data from get_random_bytes for secondary seeding
* don't reduce state space on srandom32()
* enforce state variable initialization restrictions
Note: the second paper has a version of random32() with even longer period
and a version of random64() if needed.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In the old acer_acpi, I discovered that on some of the newer AMW0 laptops
that supported the WMID methods, they don't work properly for setting the
wireless and bluetooth values.
So for the AMW0 V2 laptops, we want to use both the 'old' AMW0 and the
'new' WMID methods for setting wireless & bluetooth to guarantee we always
enable it.
This was fixed in acer_acpi some time ago, but I forgot to port the patch
over to acer-wmi when it was merged.
(Without this patch, early AMW0 V2 laptops such as the Aspire 5040 won't
work with acer-wmi, where-as they did with the old acer_acpi).
There is a slight chance for a deadlock in the estimator code. We can't call
del_timer_sync() while holding our lock, as the timer might be active and
spinning for the lock on another cpu. Work around this issue by using
try_to_del_timer_sync() and releasing the lock. We could actually delete the
timer outside of our lock, as the add and kill functions are only every called
from userspace via [gs]etsockopt() and are serialized by a mutex, but better
make this explicit.
Signed-off-by: Sven Wegener <sven.wegener@stealer.net> Acked-by: Simon Horman <horms@verge.net.au> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1121) fixes a bug in the USB serial core. When a device
is unregistered, the core will give back its minors -- even if the
device hasn't been assigned any!
The patch reserves the highest minor value (255) to mean that no minor
was assigned. It also removes some dead code and does a small style
fixup.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
usb-storage: quirk around v1.11 firmware on Nikon D40
https://bugzilla.redhat.com/show_bug.cgi?id=454028
Just as in earlier firmware versions, we need to perform this
quirk for the latest version too.
Speculatively do the entry for the D80 too, as they seem to
have the same firmware problems historically.
Signed-off-by: Dave Jones <davej@redhat.com> Cc: Johannes Berg <johannes@sipsolutions.net> Cc: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1115) adds unusual_devs entries with the IGNORE_RESIDE
flag for the iRiver T10 and the Simple Tech/Datafab CF+SM card
reader. Apparently these devices provide reasonable residue values
for READ and WRITE operations, but not for others like INQUIRY or READ
CAPACITY.
This fixes the iRiver T10 problem reported in Bugzilla #11125.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1122) fixes a bug: When an interface is unregistered,
its children (sysfs files and endpoint devices) are unregistered after
it instead of before.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Tested-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The machine will crash if the i2c_attach_client() or maven_init_client()
calls fail, although nobody has yet reported this happening.
Signed-off-by: Jean Delvare <khali@linux-fr.org> Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl> Cc: Petr Vandrovec <VANDROVE@vc.cvut.cz> 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 chips appear to have the 2D engine hang during screen redraw,
typically in a sequence of copyarea operations. This appear to be
solved by adding a flush of the engine destination pixel cache
and waiting for the engine to be idle before issuing the accel
operation. The performance impact seems to be fairly small.
Here is a trace on an RV370 (PCI device ID 0x5b64), it records the
RBBM_STATUS register, then the source x/y, destination x/y, and
width/height used for the copy:
When things are going fine the copies complete before the next ROP is
even issued, but all of a sudden the 2D unit becomes active (bit 17 in
RBBM_STATUS) and the FIFO retry (bit 13) and FIFO pipeline busy (bit
14) are set as well. The FIFO begins to backup until it becomes full.
What happens next is the radeon_fifo_wait() times out, and we access
the chip illegally leading to a bus error which usually wedges the
box. None of this makes it to the console screen, of course :-)
radeon_fifo_wait() should be modified to reset the accelerator when
this timeout happens instead of programming the chip anyways.
Another quirk is that these copyarea calls will not happen until the
first drivers/char/vt.c:redraw_screen() occurs. This will only happen
if you 1) VC switch or 2) run "consolechars" or 3) unblank the screen.
This seems to happen because until a redraw_screen() the screen scrolling
method used by fbcon is not finalized yet. I've seen this with other fb
drivers too.
So if all you do is boot straight into X you will never see this bug on
the relevant chips.
Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.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>
The bug was reported and analysed by Mark McLoughlin <markmc@redhat.com>,
the patch is based on his and Roland's suggestions.
posix_timer_event() always rewrites the pre-allocated siginfo before sending
the signal. Most of the written info is the same all the time, but memset(0)
is very wrong. If ->sigq is queued we can race with collect_signal() which
can fail to find this siginfo looking at .si_signo, or copy_siginfo() can
copy the wrong .si_code/si_tid/etc.
In short, sys_timer_settime() can in fact stop the active timer, or the user
can receive the siginfo with the wrong .si_xxx values.
Move "memset(->info, 0)" from posix_timer_event() to alloc_posix_timer(),
change send_sigqueue() to set .si_overrun = 0 when ->sigq is not queued.
It would be nice to move the whole sigq->info initialization from send to
create path, but this is not easy to do without uglifying timer_create()
further.
As Roland rightly pointed out, we need more cleanups/fixes here, see the
"FIXME" comment in the patch. Hopefully this patch makes sense anyway, and
it can mask the most bad implications.
Reported-by: Mark McLoughlin <markmc@redhat.com> Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Mark McLoughlin <markmc@redhat.com> Cc: Oliver Pinter <oliver.pntr@gmail.com> Cc: Roland McGrath <roland@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Aesthetic regards aside, commit e8e7b9eb11c34ee18bde8b7011af41938d1ad667
still leaves a bug in the error message, because it uses the unconverted
big-endian value for printk.
Fix this by using a local variable in machine byte order. The result is
correct, more readable, and also produces slightly shorter code on i386.
Signed-off-by: Petr Tesarik <ptesarik@suse.cz> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: <stable@kernel.org> Acked-by: Borislav Petkov <petkovbb@gmail.com>
[bart: __u32 -> u32] Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In relay's current read implementation, if the buffer is completely full
but hasn't triggered the buffer-full condition (i.e. the last write
didn't cross the subbuffer boundary) and the last subbuffer is exactly
full, the subbuffer accounting code erroneously finds nothing available.
This patch fixes the problem.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com> Cc: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro> Cc: Pekka Enberg <penberg@cs.helsinki.fi> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Mathieu Desnoyers <compudj@krystal.dyndns.org> Cc: Andrea Righi <righi.andrea@gmail.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>
Signed-off-by: HighPoint Linux Team <linux@highpoint-tech.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
It seems cdrwtool in the udftools has been unusable on "modern" kernels
for some time. A Google search reveals many people with the same issue
but no solution (cdrwtool fails to format the disk). After spending some
time tracking down the issue, it comes down to the following:
The udftools still use the older CDROM_SEND_PACKET interface to send
things like FORMAT_UNIT through to the drive. They should really be
updated, but that's another story. Since most distros are using libata
now, the cd or dvd burner appears as a SCSI device, and we wind up in
block/scsi_ioctl.c. Here, the code tries to take the "struct
cdrom_generic_command" and translate it and stuff it into a "struct
sg_io_hdr" structure so it can pass it to the modern sg_io() routine
instead. Unfortunately, there is one error, or rather an omission in the
translation. The timeout that is passed in in the "struct
cdrom_generic_command" is in HZ=100 units, and this is modified and
correctly converted to jiffies by use of clock_t_to_jiffies(). However,
a little further down, this cgc.timeout value in jiffies is simply
copied into the sg_io_hdr timeout, which should be in milliseconds.
Since most modern x86 kernels seems to be getting build with HZ=250, the
timeout that is passed to sg_io and eventually converted to the
timeout_per_command member of the scsi_cmnd structure is now four times
too small. Since cdrwtool tries to set the timeout to one hour for the
FORMAT_UNIT command, and it takes about 20 minutes to format a 4x CDRW,
the SCSI error-handler kicks in after the FORMAT_UNIT completes because
it took longer than the incorrectly-calculated timeout.
[jejb: fix up whitespace] Signed-off-by: Tim Wright <timw@splhi.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The class_device->device conversion is causing an oops in revalidate
because it's assuming that the device_for_each_child iterator will only
return struct scsi_device children. The conversion made all former
class_devices children of the device as well, so this assumption is
broken. Fix it.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
That seem to imply we're running off the end of the VPD inquiry data
(although at 512 bytes, it should be long enough for just about
anything). we should be using correctly sized buffers anyway, so put
those in and hope this oops goes away.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
addr = 0;
while (1)
{
/* map a page into memory each time*/
if ((addr = (char *) mmap(addr,page_size, PROT_READ |
PROT_WRITE,MAP_SHARED,fd,0)) == MAP_FAILED)
{
printf("cant do mmap on file\n");
exit(1);
}
if (0 == i)
addr1 = addr;
i++;
errno = 0;
/* lock the mapped memory pagewise*/
if ((ret = mlock((char *)addr, 1500)) == -1)
{
printf("errno value is %d\n", errno);
printf("cant lock maped region\n");
exit(1);
}
addr = addr + page_size;
}
}
======================================================
This testcase results in an mlock() failure with errno 14 that is EFAULT,
but it has nowhere been specified that mlock() will return EFAULT. When I
tested the same on older kernels like 2.6.18, I got the correct result i.e
errno 12 (ENOMEM).
I think in source code mlock(2), setting errno ENOMEM has been missed in
do_mlock() , on mlock_fixup() failure.
SUSv3 requires the following behavior frmo mlock(2).
[ENOMEM]
Some or all of the address range specified by the addr and
len arguments does not correspond to valid mapped pages
in the address space of the process.
[EAGAIN]
Some or all of the memory identified by the operation could not
be locked when the call was made.
This rule isn't so nice and slighly strange. but many people think
POSIX/SUS compliance is important.
snd_seq_oss_synth_make_info() incorrectly reports information
to userspace without first checking for the validity of the
device number, leading to possible information leak (CVE-2008-3272).
Don't forget to kill tasklets on stop to not panic if they
fire after freeing some structures.
Signed-off-by: Jiri Slaby <jirislaby@gmail.com> Acked-by: Nick Kossifidis <mickflemm@gmail.com> Cc: Luis R. Rodriguez <mcgrof@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When signal is noisy, hardware can use all RX buffers and since the last
entry in the list is self-linked, it overwrites the entry until we link
new buffers.
Ensure that we don't free this last one until we are 100% sure that it
is not used by the hardware anymore to not cause memory curruption as
can be seen below.
This is done by checking next buffer in the list. Even after that we
know that the hardware refetched the new link and proceeded further
(the next buffer is ready) we can finally free the overwritten buffer.
We discard it since the status in its descriptor is overwritten (OR-ed
by new status) too.
Signed-off-by: Jiri Slaby <jirislaby@gmail.com> Acked-by: Nick Kossifidis <mickflemm@gmail.com> Cc: Luis R. Rodriguez <mcgrof@gmail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Lookup can install a child dentry for a deleted directory. This keeps
the directory dentry alive, and the inode pinned in the cache and on
disk, even after all external references have gone away.
This isn't a big problem normally, since memory pressure or umount
will clear out the directory dentry and its children, releasing the
inode. But for UBIFS this causes problems because its orphan area can
overflow.
Fix this by returning ENOENT for all lookups on a S_DEAD directory
before creating a child dentry.
Thanks to Zoltan Sogor for noticing this while testing UBIFS, and
Artem for the excellent analysis of the problem and testing.
projects / karo-tx-linux.git / log