firefly-linux-kernel-4.4.55.git
18 years ago[TCP]: Use TCPOLEN_TSTAMP_ALIGNED macro instead of magic number.
YOSHIFUJI Hideaki [Wed, 11 Oct 2006 02:40:50 +0000 (19:40 -0700)]
[TCP]: Use TCPOLEN_TSTAMP_ALIGNED macro instead of magic number.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IPV6]: Seperate sit driver to extra module (addrconf.c changes)
Joerg Roedel [Tue, 10 Oct 2006 21:49:53 +0000 (14:49 -0700)]
[IPV6]: Seperate sit driver to extra module (addrconf.c changes)

This patch contains the changes to net/ipv6/addrconf.c to remove sit
specific code if the sit driver is not selected.

Signed-off-by: Joerg Roedel <joro-lkml@zlug.org>
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IPV6]: Seperate sit driver to extra module
Joerg Roedel [Tue, 10 Oct 2006 21:47:44 +0000 (14:47 -0700)]
[IPV6]: Seperate sit driver to extra module

This patch removes the driver of the IPv6-in-IPv4 tunnel driver (sit)
from the IPv6 module. It adds an option to Kconfig which makes it
possible to compile it as a seperate module.

Signed-off-by: Joerg Roedel <joro-lkml@zlug.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NET]: File descriptor loss while receiving SCM_RIGHTS
Miklos Szeredi [Tue, 10 Oct 2006 04:42:14 +0000 (21:42 -0700)]
[NET]: File descriptor loss while receiving SCM_RIGHTS

If more than one file descriptor was sent with an SCM_RIGHTS message,
and on the receiving end, after installing a nonzero (but not all)
file descritpors the process runs out of fds, then the already
installed fds will be lost (userspace will have no way of knowing
about them).

The following patch makes sure, that at least the already installed
fds are sent to userspace.  It doesn't solve the issue of losing file
descriptors in case of an EFAULT on the userspace buffer.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[SCTP]: Fix the RX queue size shown in /proc/net/sctp/assocs output.
Vlad Yasevich [Tue, 10 Oct 2006 04:34:26 +0000 (21:34 -0700)]
[SCTP]: Fix the RX queue size shown in /proc/net/sctp/assocs output.

Show the true receive buffer usage.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[SCTP]: Fix receive buffer accounting.
Vlad Yasevich [Tue, 10 Oct 2006 04:34:04 +0000 (21:34 -0700)]
[SCTP]: Fix receive buffer accounting.

When doing receiver buffer accounting, we always used skb->truesize.
This is problematic when processing bundled DATA chunks because for
every DATA chunk that could be small part of one large skb, we would
charge the size of the entire skb.  The new approach is to store the
size of the DATA chunk we are accounting for in the sctp_ulpevent
structure and use that stored value for accounting.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years agoSELinux: Bug fix in polidydb_destroy
Chad Sellers [Fri, 6 Oct 2006 20:09:52 +0000 (16:09 -0400)]
SELinux: Bug fix in polidydb_destroy

This patch fixes two bugs in policydb_destroy. Two list pointers
(policydb.ocontexts[i] and policydb.genfs) were not being reset to NULL when
the lists they pointed to were being freed. This caused a problem when the
initial policy load failed, as the policydb being destroyed was not a
temporary new policydb that was thrown away, but rather was the global
(active) policydb. Consequently, later functions, particularly
sys_bind->selinux_socket_bind->security_node_sid and
do_rw_proc->selinux_sysctl->selinux_proc_get_sid->security_genfs_sid tried
to dereference memory that had previously been freed.

Signed-off-by: Chad Sellers <csellers@tresys.com>
Signed-off-by: James Morris <jmorris@namei.org>
18 years agoIPsec: fix handling of errors for socket policies
Venkat Yekkirala [Thu, 5 Oct 2006 20:42:35 +0000 (15:42 -0500)]
IPsec: fix handling of errors for socket policies

This treats the security errors encountered in the case of
socket policy matching, the same as how these are treated in
the case of main/sub policies, which is to return a full lookup
failure.

Signed-off-by: Venkat Yekkirala <vyekkirala@TrustedCS.com>
Signed-off-by: James Morris <jmorris@namei.org>
18 years agoIPsec: correct semantics for SELinux policy matching
Venkat Yekkirala [Thu, 5 Oct 2006 20:42:18 +0000 (15:42 -0500)]
IPsec: correct semantics for SELinux policy matching

Currently when an IPSec policy rule doesn't specify a security
context, it is assumed to be "unlabeled" by SELinux, and so
the IPSec policy rule fails to match to a flow that it would
otherwise match to, unless one has explicitly added an SELinux
policy rule allowing the flow to "polmatch" to the "unlabeled"
IPSec policy rules. In the absence of such an explicitly added
SELinux policy rule, the IPSec policy rule fails to match and
so the packet(s) flow in clear text without the otherwise applicable
xfrm(s) applied.

The above SELinux behavior violates the SELinux security notion of
"deny by default" which should actually translate to "encrypt by
default" in the above case.

This was first reported by Evgeniy Polyakov and the way James Morris
was seeing the problem was when connecting via IPsec to a
confined service on an SELinux box (vsftpd), which did not have the
appropriate SELinux policy permissions to send packets via IPsec.

With this patch applied, SELinux "polmatching" of flows Vs. IPSec
policy rules will only come into play when there's a explicit context
specified for the IPSec policy rule (which also means there's corresponding
SELinux policy allowing appropriate domains/flows to polmatch to this context).

Secondly, when a security module is loaded (in this case, SELinux), the
security_xfrm_policy_lookup() hook can return errors other than access denied,
such as -EINVAL.  We were not handling that correctly, and in fact
inverting the return logic and propagating a false "ok" back up to
xfrm_lookup(), which then allowed packets to pass as if they were not
associated with an xfrm policy.

The solution for this is to first ensure that errno values are
correctly propagated all the way back up through the various call chains
from security_xfrm_policy_lookup(), and handled correctly.

Then, flow_cache_lookup() is modified, so that if the policy resolver
fails (typically a permission denied via the security module), the flow
cache entry is killed rather than having a null policy assigned (which
indicates that the packet can pass freely).  This also forces any future
lookups for the same flow to consult the security module (e.g. SELinux)
for current security policy (rather than, say, caching the error on the
flow cache entry).

This patch: Fix the selinux side of things.

This makes sure SELinux polmatching of flow contexts to IPSec policy
rules comes into play only when an explicit context is associated
with the IPSec policy rule.

Also, this no longer defaults the context of a socket policy to
the context of the socket since the "no explicit context" case
is now handled properly.

Signed-off-by: Venkat Yekkirala <vyekkirala@TrustedCS.com>
Signed-off-by: James Morris <jmorris@namei.org>
18 years agoIPsec: propagate security module errors up from flow_cache_lookup
James Morris [Thu, 5 Oct 2006 20:42:27 +0000 (15:42 -0500)]
IPsec: propagate security module errors up from flow_cache_lookup

When a security module is loaded (in this case, SELinux), the
security_xfrm_policy_lookup() hook can return an access denied permission
(or other error).  We were not handling that correctly, and in fact
inverting the return logic and propagating a false "ok" back up to
xfrm_lookup(), which then allowed packets to pass as if they were not
associated with an xfrm policy.

The way I was seeing the problem was when connecting via IPsec to a
confined service on an SELinux box (vsftpd), which did not have the
appropriate SELinux policy permissions to send packets via IPsec.

The first SYNACK would be blocked, because of an uncached lookup via
flow_cache_lookup(), which would fail to resolve an xfrm policy because
the SELinux policy is checked at that point via the resolver.

However, retransmitted SYNACKs would then find a cached flow entry when
calling into flow_cache_lookup() with a null xfrm policy, which is
interpreted by xfrm_lookup() as the packet not having any associated
policy and similarly to the first case, allowing it to pass without
transformation.

The solution presented here is to first ensure that errno values are
correctly propagated all the way back up through the various call chains
from security_xfrm_policy_lookup(), and handled correctly.

Then, flow_cache_lookup() is modified, so that if the policy resolver
fails (typically a permission denied via the security module), the flow
cache entry is killed rather than having a null policy assigned (which
indicates that the packet can pass freely).  This also forces any future
lookups for the same flow to consult the security module (e.g. SELinux)
for current security policy (rather than, say, caching the error on the
flow cache entry).

Signed-off-by: James Morris <jmorris@namei.org>
18 years agoNetLabel: use SECINITSID_UNLABELED for a base SID
paul.moore@hp.com [Thu, 5 Oct 2006 22:28:24 +0000 (18:28 -0400)]
NetLabel: use SECINITSID_UNLABELED for a base SID

This patch changes NetLabel to use SECINITSID_UNLABLELED as it's source of
SELinux type information when generating a NetLabel context.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: James Morris <jmorris@namei.org>
18 years agoNetLabel: fix a cache race condition
paul.moore@hp.com [Wed, 4 Oct 2006 15:46:31 +0000 (11:46 -0400)]
NetLabel: fix a cache race condition

Testing revealed a problem with the NetLabel cache where a cached entry could
be freed while in use by the LSM layer causing an oops and other problems.
This patch fixes that problem by introducing a reference counter to the cache
entry so that it is only freed when it is no longer in use.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: James Morris <jmorris@namei.org>
18 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Wed, 11 Oct 2006 22:30:14 +0000 (15:30 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] Pass NULL not 0 for pointer value.
  [MIPS] IP27: Make declaration of setup_replication_mask a proper prototype.
  [MIPS] BigSur: More useful defconfig.
  [MIPS] Cleanup definitions of speed_t and tcflag_t.
  [MIPS] Fix compilation warnings in arch/mips/sibyte/bcm1480/smp.c
  [MIPS] Optimize and cleanup get_saved_sp, set_saved_sp
  [MIPS] <asm/irq.h> does not need pt_regs anymore.
  [MIPS] Workaround for bug in gcc -EB / -EL options.
  [MIPS] Fix timer setup for Jazz

18 years ago[MIPS] Pass NULL not 0 for pointer value.
Ralf Baechle [Wed, 11 Oct 2006 18:30:03 +0000 (19:30 +0100)]
[MIPS] Pass NULL not 0 for pointer value.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] IP27: Make declaration of setup_replication_mask a proper prototype.
Ralf Baechle [Wed, 11 Oct 2006 17:35:33 +0000 (18:35 +0100)]
[MIPS] IP27: Make declaration of setup_replication_mask a proper prototype.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] BigSur: More useful defconfig.
Ralf Baechle [Wed, 11 Oct 2006 00:42:30 +0000 (01:42 +0100)]
[MIPS] BigSur: More useful defconfig.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] Cleanup definitions of speed_t and tcflag_t.
Ralf Baechle [Wed, 11 Oct 2006 00:22:12 +0000 (01:22 +0100)]
[MIPS] Cleanup definitions of speed_t and tcflag_t.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] Fix compilation warnings in arch/mips/sibyte/bcm1480/smp.c
Mark Mason [Wed, 11 Oct 2006 00:03:10 +0000 (17:03 -0700)]
[MIPS] Fix compilation warnings in arch/mips/sibyte/bcm1480/smp.c

Signed-off-by: Mark Mason <mason@broadcom.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] Optimize and cleanup get_saved_sp, set_saved_sp
Atsushi Nemoto [Tue, 10 Oct 2006 13:46:52 +0000 (22:46 +0900)]
[MIPS] Optimize and cleanup get_saved_sp, set_saved_sp

If CONFIG_BUILD_ELF64 was not selected and gcc had -msym32 option
(i.e. 4.0 or newer), there is no point to use %highest, %higher for
kernel symbols.

This patch also fixes 64-bit SMTC version of get_saved_sp() which is
broken but harmless since there is no such CPUs for now.

A bonus is set_saved_sp() and SMP version of get_saved_sp() are more
readable now.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] <asm/irq.h> does not need pt_regs anymore.
Atsushi Nemoto [Tue, 10 Oct 2006 16:07:01 +0000 (01:07 +0900)]
[MIPS] <asm/irq.h> does not need pt_regs anymore.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] Workaround for bug in gcc -EB / -EL options.
Ralf Baechle [Tue, 10 Oct 2006 14:44:10 +0000 (15:44 +0100)]
[MIPS] Workaround for bug in gcc -EB / -EL options.

Certain gcc versions upto gcc 4.1.1 (probably 4.2-subversion as of
2006-10-10 don't properly change the the predefined symbols if -EB / -EL
are used, so we kludge that here.  A bug has been filed at
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29413.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years ago[MIPS] Fix timer setup for Jazz
Yoichi Yuasa [Tue, 10 Oct 2006 07:56:11 +0000 (16:56 +0900)]
[MIPS] Fix timer setup for Jazz

Signed-off-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
18 years agoMerge branch 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Wed, 11 Oct 2006 18:20:11 +0000 (11:20 -0700)]
Merge branch 'for-linus' of git://git390.osdl.marist.edu/linux-2.6

* 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [S390] stacktrace bug.
  [S390] cio: remove casts from/to (void *).
  [S390] cio: Remove grace period for vary off chpid.
  [S390] cio: Use ccw_dev_id and subchannel_id in ccw_device_private
  [S390] monwriter kzalloc size.
  [S390] cio: add missing KERN_INFO printk header.
  [S390] irq change improvements.

18 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Wed, 11 Oct 2006 18:19:47 +0000 (11:19 -0700)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  [PATCH] pata-qdi: fix le32 in data_xfer
  [libata] sata_promise: add PCI ID
  [PATCH] libata: return sense data in HDIO_DRIVE_CMD ioctl
  [PATCH] libata: Don't believe bogus claims in the older PIO mode register

18 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Wed, 11 Oct 2006 18:19:30 +0000 (11:19 -0700)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/netdev-2.6

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6: (42 commits)
  [PATCH] Fix section mismatch in de2104x.c
  [PATCH] sky2: set lower pause threshold to prevent overrun
  [PATCH] sky2: revert pci express extensions
  [PATCH] skge: version 1.9
  [PATCH] skge: better flow control negotiation
  [PATCH] skge: pause mapping for fiber
  [PATCH] skge: fix stuck irq when fiber down
  [PATCH] powerpc/cell spidernet release all descrs
  [PATCH] powerpc/cell spidernet DMA direction fix
  [PATCH] powerpc/cell spidernet variable name change
  [PATCH] powerpc/cell spidernet reduce DMA kicking
  [PATCH] powerpc/cell spidernet
  [PATCH] powerpc/cell spidernet refine locking
  [PATCH] powerpc/cell spidernet NAPI polling info.
  [PATCH] powerpc/cell spidernet low watermark patch.
  [PATCH] powerpc/cell spidernet incorrect offset
  [PATCH] powerpc/cell spidernet stop error printing patch.
  [PATCH] powerpc/cell spidernet fix error interrupt print
  [PATCH] powerpc/cell spidernet bogus rx interrupt bit
  [PATCH] Spidernet stop queue when queue is full.
  ...

18 years ago[PATCH] i2Output always takes kernel data now
Al Viro [Wed, 11 Oct 2006 16:45:47 +0000 (17:45 +0100)]
[PATCH] i2Output always takes kernel data now

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] alpha_ksyms.c cleanup
Al Viro [Wed, 11 Oct 2006 16:40:22 +0000 (17:40 +0100)]
[PATCH] alpha_ksyms.c cleanup

taken exports to actual definitions of symbols being exported.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] remove bogus arch-specific syscall exports
Al Viro [Wed, 11 Oct 2006 16:40:22 +0000 (17:40 +0100)]
[PATCH] remove bogus arch-specific syscall exports

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] z2_init() in non-modular case
Al Viro [Wed, 11 Oct 2006 16:28:47 +0000 (17:28 +0100)]
[PATCH] z2_init() in non-modular case

... another victim - this time of 2.5.1-pre2

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] amiga_floppy_init() in non-modular case
Al Viro [Wed, 11 Oct 2006 16:28:37 +0000 (17:28 +0100)]
[PATCH] amiga_floppy_init() in non-modular case

It used to be called directly, but that got lost in 2.1.87-pre1.
Similar breakage in ataflop got fixed 3 years ago, this one
had gone unnoticed.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] clean m68k ksyms
Al Viro [Wed, 11 Oct 2006 16:28:27 +0000 (17:28 +0100)]
[PATCH] clean m68k ksyms

sun3_ksyms gone, m68k_ksyms trimmed down to exports of the assembler ones,
for sun3 added the missing exports of __ioremap() and iounmap().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] sun3 __iomem annotations
Al Viro [Wed, 11 Oct 2006 16:28:17 +0000 (17:28 +0100)]
[PATCH] sun3 __iomem annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] misc m68k __user annotations
Al Viro [Wed, 11 Oct 2006 16:28:07 +0000 (17:28 +0100)]
[PATCH] misc m68k __user annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68k uaccess __user annotations
Al Viro [Wed, 11 Oct 2006 16:27:57 +0000 (17:27 +0100)]
[PATCH] m68k uaccess __user annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] misuse of strstr
Al Viro [Wed, 11 Oct 2006 16:25:45 +0000 (17:25 +0100)]
[PATCH] misuse of strstr

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m32r: more __user annotations
Al Viro [Wed, 11 Oct 2006 16:24:55 +0000 (17:24 +0100)]
[PATCH] m32r: more __user annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m32r: NULL noise removal
Al Viro [Wed, 11 Oct 2006 16:24:35 +0000 (17:24 +0100)]
[PATCH] m32r: NULL noise removal

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m32r: signal __user annotations
Al Viro [Wed, 11 Oct 2006 16:24:45 +0000 (17:24 +0100)]
[PATCH] m32r: signal __user annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m32r: C99 initializers in setup.c
Al Viro [Wed, 11 Oct 2006 16:24:25 +0000 (17:24 +0100)]
[PATCH] m32r: C99 initializers in setup.c

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] arm-versatile iomem annotations
Al Viro [Wed, 11 Oct 2006 16:22:34 +0000 (17:22 +0100)]
[PATCH] arm-versatile iomem annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] arm: use unsigned long instead of unsigned int in get_user()
Al Viro [Wed, 11 Oct 2006 16:22:54 +0000 (17:22 +0100)]
[PATCH] arm: use unsigned long instead of unsigned int in get_user()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] arm __user annotations
Al Viro [Wed, 11 Oct 2006 16:22:44 +0000 (17:22 +0100)]
[PATCH] arm __user annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] md: use BUILD_BUG_ON
Alexey Dobriyan [Wed, 11 Oct 2006 08:22:26 +0000 (01:22 -0700)]
[PATCH] md: use BUILD_BUG_ON

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ISDN: several minor fixes
Jeff Garzik [Wed, 11 Oct 2006 08:22:25 +0000 (01:22 -0700)]
[PATCH] ISDN: several minor fixes

pcbit: kill 'may be used uninitialized' warning.  although the code does
eventually fill the 32 bits it cares about, the variable truly is
accessed uninitialized in each macro.  Easier to just clean it up now.

sc: fix a ton of obviously incorrect printk's (some with missing
arguments even)

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] fs/bio.c: tweaks
Andreas Mohr [Wed, 11 Oct 2006 08:22:24 +0000 (01:22 -0700)]
[PATCH] fs/bio.c: tweaks

- Calculate a variable in bvec_alloc_bs() only once needed, not earlier
  (bio.o down from 18408 to 18376 Bytes, 32 Bytes saved, probably due to
  data locality improvements).

- Init variable idx to silence a gcc warning which already existed in the
  unmodified original base file (bvec_alloc_bs() handles idx correctly, so
  there's no need for the warning):

fs/bio.c: In function `bio_alloc_bioset':
fs/bio.c:169: warning: `idx' may be used uninitialized in this function

Signed-off-by: Andreas Mohr <andi@lisas.de>
Acked-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] drivers/mca: handle sysfs errors
Jeff Garzik [Wed, 11 Oct 2006 08:22:23 +0000 (01:22 -0700)]
[PATCH] drivers/mca: handle sysfs errors

Also includes a kmalloc->kzalloc cleanup.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] firmware/efivars: handle error
Jeff Garzik [Wed, 11 Oct 2006 08:22:23 +0000 (01:22 -0700)]
[PATCH] firmware/efivars: handle error

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] EISA: handle sysfs errors
Jeff Garzik [Wed, 11 Oct 2006 08:22:22 +0000 (01:22 -0700)]
[PATCH] EISA: handle sysfs errors

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ipmi: handle sysfs errors
Jeff Garzik [Wed, 11 Oct 2006 08:22:21 +0000 (01:22 -0700)]
[PATCH] ipmi: handle sysfs errors

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Corey Minyard <cminyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] firmware/dell_rbu: handle sysfs errors
Jeff Garzik [Wed, 11 Oct 2006 08:22:20 +0000 (01:22 -0700)]
[PATCH] firmware/dell_rbu: handle sysfs errors

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Cc: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] VFS: Destroy the dentries contributed by a superblock on unmounting
David Howells [Wed, 11 Oct 2006 08:22:19 +0000 (01:22 -0700)]
[PATCH] VFS: Destroy the dentries contributed by a superblock on unmounting

The attached patch destroys all the dentries attached to a superblock in one go
by:

 (1) Destroying the tree rooted at s_root.

 (2) Destroying every entry in the anon list, one at a time.

 (3) Each entry in the anon list has its subtree consumed from the leaves
     inwards.

This reduces the amount of work generic_shutdown_super() does, and avoids
iterating through the dentry_unused list.

Note that locking is almost entirely absent in the shrink_dcache_for_umount*()
functions added by this patch.  This is because:

 (1) at the point the filesystem calls generic_shutdown_super(), it is not
     permitted to further touch the superblock's set of dentries, and nor may
     it remove aliases from inodes;

 (2) the dcache memory shrinker now skips dentries that are being unmounted;
     and

 (3) the superblock no longer has any external references through which the VFS
     can reach it.

Given these points, the only locking we need to do is when we remove dentries
from the unused list and the name hashes, which we do a directory's worth at a
time.

We also don't need to guard against reference counts going to zero unexpectedly
and removing bits of the tree we're working on as nothing else can call dput().

A cut down version of dentry_iput() has been folded into
shrink_dcache_for_umount_subtree() function.  Apart from not needing to unlock
things, it also doesn't need to check for inotify watches.

In this version of the patch, the complaint about a dentry still being in use
has been expanded from a single BUG_ON() and now gives much more information.

Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: NeilBrown <neilb@suse.de>
Acked-by: Ian Kent <raven@themaw.net>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AUTOFS: Make sure all dentries refs are released before calling kill_anon_super()
David Howells [Wed, 11 Oct 2006 08:22:15 +0000 (01:22 -0700)]
[PATCH] AUTOFS: Make sure all dentries refs are released before calling kill_anon_super()

Make sure all dentries refs are released before calling kill_anon_super() so
that the assumption that generic_shutdown_super() can completely destroy the
dentry tree for there will be no external references holds true.

What was being done in the put_super() superblock op, is now done in the
kill_sb() filesystem op instead, prior to calling kill_anon_super().

This makes the struct autofs_sb_info::root member variable redundant (since
sb->s_root is still available), and so that is removed.  The calls to
shrink_dcache_sb() are also removed since they're also redundant as
shrink_dcache_for_umount() will now be called after the cleanup routine.

Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Ian Kent <raven@themaw.net>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ReiserFS: Make sure all dentries refs are released before calling kill_block_...
David Howells [Wed, 11 Oct 2006 08:22:14 +0000 (01:22 -0700)]
[PATCH] ReiserFS: Make sure all dentries refs are released before calling kill_block_super()

Make sure all dentries refs are released before calling kill_block_super()
so that the assumption that generic_shutdown_super() can completely destroy
the dentry tree for there will be no external references holds true.

What was being done in the put_super() superblock op, is now done in the
kill_sb() filesystem op instead, prior to calling kill_block_super().

Changes made in [try #2]:

 (*) reiserfs_kill_sb() now checks that the superblock FS info pointer is set
     before trying to dereference it.

Signed-off-by: David Howells <dhowells@redhat.com>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: <reiserfs-dev@namesys.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] dell_rbu: printk() warning fix
Andrew Morton [Wed, 11 Oct 2006 08:22:13 +0000 (01:22 -0700)]
[PATCH] dell_rbu: printk() warning fix

drivers/firmware/dell_rbu.c: In function 'packetize_data':
drivers/firmware/dell_rbu.c:252: warning: format '%lu' expects type 'long unsigned int', but argument 3 has type 'int'

Cc: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] kernel-doc: make parameter description indentation uniform
Randy Dunlap [Wed, 11 Oct 2006 08:22:12 +0000 (01:22 -0700)]
[PATCH] kernel-doc: make parameter description indentation uniform

- In parameter descriptions, strip all whitespace between the parameter
  name (e.g., @len) and its description so that the description is
  indented uniformly in text and man page modes.  Previously, spaces
  or tabs (which are used for cleaner source code viewing) affected
  the produced output in a negative way.

Before (man mode):
       to            Destination address, in user space.
       from        Source address, in kernel space.
       n              Number of bytes to copy.

After (man mode):
       to          Destination address, in user space.
       from        Source address, in kernel space.
       n           Number of bytes to copy.

- Fix/clarify a few function description comments.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] include linux/types.h in linux/nbd.h
Mike Frysinger [Wed, 11 Oct 2006 08:22:11 +0000 (01:22 -0700)]
[PATCH] include linux/types.h in linux/nbd.h

The nbd header uses __be32 and such types but doesn't actually include the
header that defines these things (linux/types.h); so let's include it.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] kernel-doc: drop various "inline" qualifiers
Randy Dunlap [Wed, 11 Oct 2006 08:22:10 +0000 (01:22 -0700)]
[PATCH] kernel-doc: drop various "inline" qualifiers

Drop __inline, __always_inline, and noinline in the produced kernel-doc
output, similar to other pseudo directives.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uaccess.h: match kernel-doc and function names
Randy Dunlap [Wed, 11 Oct 2006 08:22:10 +0000 (01:22 -0700)]
[PATCH] uaccess.h: match kernel-doc and function names

Place kernel-doc function comment header immediately before the function that
is being documented.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] kernel-doc: fix function name in usercopy.c
Randy Dunlap [Wed, 11 Oct 2006 08:22:09 +0000 (01:22 -0700)]
[PATCH] kernel-doc: fix function name in usercopy.c

Fix kernel-doc function name in usercopy.c.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] lockdep: fix printk recursion logic
Ingo Molnar [Wed, 11 Oct 2006 08:22:08 +0000 (01:22 -0700)]
[PATCH] lockdep: fix printk recursion logic

Bug reported and fixed by Tilman Schmidt <tilman@imap.cc>: if lockdep is
enabled then log messages make it to /var/log/messages belatedly.  The
reason is a missed wakeup of klogd.

Initially there was only a lockdep_internal() protection against lockdep
recursion within vprintk() - it grew the 'outer' lockdep_off()/on()
protection only later on.  But that lockdep_off() made the
release_console_sem() within vprintk() always happen under the
lockdep_internal() condition, causing the bug.

The right solution to remove the inner protection against recursion here -
the outer one is enough.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] fix lockdep-design.txt
Aneesh Kumar [Wed, 11 Oct 2006 08:22:07 +0000 (01:22 -0700)]
[PATCH] fix lockdep-design.txt

I was looking at lockdep-desing.txt and i guess i am confused with the
changes with respect to fd7bcea35e7efb108c34ee2b3840942a3749cadb. It
says

+   '.'  acquired while irqs enabled
+   '+'  acquired in irq context
+   '-'  acquired in process context with irqs disabled
+   '?'  read-acquired both with irqs enabled and in irq context
+

But the get_usage_chars() function does this for '-'
 if (class->usage_mask & LOCKF_ENABLED_HARDIRQS)
                        *c1 = '-';

So i guess what would be correct would be
'.'  acquired while irqs disabled
'+'  acquired in irq context
'-'  acquired with irqs enabled
'?' read acquired in irq context with irqs enabled.

Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] lockdep: use BUILD_BUG_ON
Alexey Dobriyan [Wed, 11 Oct 2006 08:22:06 +0000 (01:22 -0700)]
[PATCH] lockdep: use BUILD_BUG_ON

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] DAC960: use memmove for overlapping areas
Alexey Dobriyan [Wed, 11 Oct 2006 08:22:06 +0000 (01:22 -0700)]
[PATCH] DAC960: use memmove for overlapping areas

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] fs/*: use BUILD_BUG_ON
Alexey Dobriyan [Wed, 11 Oct 2006 08:22:05 +0000 (01:22 -0700)]
[PATCH] fs/*: use BUILD_BUG_ON

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Mark Fasheh <mark.fasheh@oracle.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] HT_IRQ must depend on PCI
Adrian Bunk [Wed, 11 Oct 2006 08:22:04 +0000 (01:22 -0700)]
[PATCH] HT_IRQ must depend on PCI

CONFIG_PCI=n, CONFIG_HT_IRQ=y results in the following compile error:

...
  LD      vmlinux
arch/i386/mach-generic/built-in.o: In function `apicid_to_node':
summit.c:(.text+0x53): undefined reference to `apicid_2_node'
arch/i386/kernel/built-in.o: In function `arch_setup_ht_irq':
(.text+0xcf79): undefined reference to `write_ht_irq_low'
arch/i386/kernel/built-in.o: In function `arch_setup_ht_irq':
(.text+0xcf85): undefined reference to `write_ht_irq_high'
arch/i386/kernel/built-in.o: In function `k7nops':
alternative.c:(.data+0x1358): undefined reference to `mask_ht_irq'
alternative.c:(.data+0x1360): undefined reference to `unmask_ht_irq'
make[1]: *** [vmlinux] Error 1

Bug report by Jesper Juhl.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Fix typos in mm/shmem_acl.c
Aneesh Kumar [Wed, 11 Oct 2006 08:22:03 +0000 (01:22 -0700)]
[PATCH] Fix typos in mm/shmem_acl.c

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Consolidate check_signature
Matthew Wilcox [Wed, 11 Oct 2006 08:22:02 +0000 (01:22 -0700)]
[PATCH] Consolidate check_signature

There's nothing arch-specific about check_signature(), so move it to
<linux/io.h>.  Use a cross between the Alpha and i386 implementations as
the generic one.

Signed-off-by: Matthew Wilcox <willy@parisc-linux.org>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Use linux/io.h instead of asm/io.h
Matthew Wilcox [Wed, 11 Oct 2006 08:22:01 +0000 (01:22 -0700)]
[PATCH] Use linux/io.h instead of asm/io.h

In preparation for moving check_signature, change these users from asm/io.h
to linux/io.h

Signed-off-by: Matthew Wilcox <willy@parisc-linux.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] D-cache aliasing issue in __block_prepare_write
Monakhov Dmitriy [Wed, 11 Oct 2006 08:22:00 +0000 (01:22 -0700)]
[PATCH] D-cache aliasing issue in __block_prepare_write

A couple of flush_dcache_page()s are missing on the I/O-error paths.

Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Remove unnecessary check in fs/fat/inode.c
Eric Sesterhenn [Wed, 11 Oct 2006 08:21:59 +0000 (01:21 -0700)]
[PATCH] Remove unnecessary check in fs/fat/inode.c

Aince all callers dereference sb, and this function does so earlier too, we
dont need the check.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] VM: Fix the gfp_mask in invalidate_complete_page2
Trond Myklebust [Wed, 11 Oct 2006 08:21:58 +0000 (01:21 -0700)]
[PATCH] VM: Fix the gfp_mask in invalidate_complete_page2

If try_to_release_page() is called with a zero gfp mask, then the
filesystem is effectively denied the possibility of sleeping while
attempting to release the page.  There doesn't appear to be any valid
reason why this should be banned, given that we're not calling this from a
memory allocation context.

For this reason, change the gfp_mask argument of the call to GFP_KERNEL.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Steve Dickson <SteveD@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] document the core-dump-to-a-pipe patch
Matthias Urlichs [Wed, 11 Oct 2006 08:21:57 +0000 (01:21 -0700)]
[PATCH] document the core-dump-to-a-pipe patch

The pipe-a-coredump-to-a-program feature was undocumented.
*Grumble*.

NB: a good enhancement to that patch would be: save all the stuff that a
core file can get from the %x expansions in the environment.

Signed-off-by: Matthias Urlichs <matthias@urlichs.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] bitmap: parse input from kernel and user buffers
Reinette Chatre [Wed, 11 Oct 2006 08:21:55 +0000 (01:21 -0700)]
[PATCH] bitmap: parse input from kernel and user buffers

lib/bitmap.c:bitmap_parse() is a library function that received as input a
user buffer.  This seemed to have originated from the way the write_proc
function of the /proc filesystem operates.

This has been reworked to not use kmalloc and eliminates a lot of
get_user() overhead by performing one access_ok before using __get_user().

We need to test if we are in kernel or user space (is_user) and access the
buffer differently.  We cannot use __get_user() to access kernel addresses
in all cases, for example in architectures with separate address space for
kernel and user.

This function will be useful for other uses as well; for example, taking
input for /sysfs instead of /proc, so it was changed to accept kernel
buffers.  We have this use for the Linux UWB project, as part as the
upcoming bandwidth allocator code.

Only a few routines used this function and they were changed too.

Signed-off-by: Reinette Chatre <reinette.chatre@linux.intel.com>
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Joe Korty <joe.korty@ccur.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] 32-bit compatibility HDIO IOCTLs
Maciej W. Rozycki [Wed, 11 Oct 2006 08:21:54 +0000 (01:21 -0700)]
[PATCH] 32-bit compatibility HDIO IOCTLs

A couple of HDIO IOCTLs are not yet handled and a few others are marked
as using a pointer rather than an unsigned long.  The formers include:

HDIO_GET_WCACHE, HDIO_GET_ACOUSTIC, HDIO_GET_ADDRESS and
HDIO_GET_BUSSTATE.  The latters are: HDIO_SET_MULTCOUNT,
HDIO_SET_UNMASKINTR, HDIO_SET_KEEPSETTINGS, HDIO_SET_32BIT,
HDIO_SET_NOWERR, HDIO_SET_DMA, HDIO_SET_PIO_MODE and HDIO_SET_NICE.

Additionally 0x330 used to be HDIO_GETGEO_BIG and may be issued by 32-bit
`hdparm' run on a 64-bit kernel making Linux complain loudly.

This is a fix for these issues.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86/microcode: handle sysfs error
Jeff Garzik [Wed, 11 Oct 2006 08:21:53 +0000 (01:21 -0700)]
[PATCH] x86/microcode: handle sysfs error

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] invalidate_inode_pages2_range() debug
Andrew Morton [Wed, 11 Oct 2006 08:21:53 +0000 (01:21 -0700)]
[PATCH] invalidate_inode_pages2_range() debug

A failure in invalidate_inode_pages2_range() can result in unpleasant things
happening in NFS (at least).  Stick a WARN_ON_ONCE() in there so we can find
out if it happens, and maybe why.

(akpm: might be a -mm-only patch, we'll see..)

Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Steve Dickson <SteveD@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] sched: likely profiling
Nick Piggin [Wed, 11 Oct 2006 08:21:52 +0000 (01:21 -0700)]
[PATCH] sched: likely profiling

This likely profiling is pretty fun. I found a few possible problems
in sched.c.

This patch may be not measurable, but when I did measure long ago,
nooping (un)likely cost a couple of % on scheduler heavy benchmarks, so
it all adds up.

Tweak some branch hints:

- the 2nd 64 bits in the bitmask is likely to be populated, because it
  contains the first 28 bits (nearly 3/4) of the normal priorities.
  (ratio of 669669:691 ~= 1000:1).

- it isn't unlikely that context switching switches to another process. it
  might be very rapidly switching to and from the idle process (ratio of
  475815:419004 and 471330:423544). Let the branch predictor decide.

- preempt_enable seems to be very often called in a nested preempt_disable
  or with interrupts disabled (ratio of 3567760:87965 ~= 40:1)

Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Daniel Walker <dwalker@mvista.com>
Cc: Hua Zhong <hzhong@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] tpm: fix error handling
Jeff Garzik [Wed, 11 Oct 2006 08:21:51 +0000 (01:21 -0700)]
[PATCH] tpm: fix error handling

- handle sysfs error
- handle driver model errors
- de-obfuscate platform_device_register_simple() call, which included an
  assignment in between two function calls, in the same C statement.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Acked-by: Kylene Hall <kjhall@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ext2: errors behaviour fix
Vasily Averin [Wed, 11 Oct 2006 08:21:50 +0000 (01:21 -0700)]
[PATCH] ext2: errors behaviour fix

Current error behaviour for ext2 and ext3 filesystems does not fully
correspond to the documentation and should be fixed.

According to man 8 mount, ext2 and ext3 file systems allow to set one of 3
different on-errors behaviours:

  ---- start of quote man 8 mount ----

  errors=continue / errors=remount-ro / errors=panic

    Define the behaviour when an error is encountered.  (Either ignore
    errors and just mark the file system erroneous and continue, or remount
    the file system read-only, or panic and halt the system.) The default is
    set in the filesystem superblock, and can be changed using tune2fs(8).

  ---- end of quote ----

However EXT3_ERRORS_CONTINUE is not read from the superblock, and thus
ERRORS_CONT is not saved on the sbi->s_mount_opt.  It leads to the incorrect
handle of errors on ext3.

Then we've checked corresponding code in ext2 and discovered that it is buggy
as well:

- EXT2_ERRORS_CONTINUE is not read from the superblock (the same);

- parse_option() does not clean the alternative values and thus something
  like (ERRORS_CONT|ERRORS_RO) can be set;

- if options are omitted, parse_option() does not set any of these options.

Therefore it is possible to set any combination of these options on the ext2:

- none of them may be set: EXT2_ERRORS_CONTINUE on superblock / empty mount
  options;

- any of them may be set using mount options;

- 2 any options may be set: by using EXT2_ERRORS_RO/EXT2_ERRORS_PANIC on the
  superblock and other value in mount options;

- and finally all three options may be set by adding third option in remount.

Currently ext2 uses these values only in ext2_error() and it is not leading to
any noticeable troubles.  However somebody may be discouraged when he will try
to workaround EXT2_ERRORS_PANIC on the superblock by using errors=continue in
mount options.

This patch:

EXT2_ERRORS_CONTINUE should be read from the superblock as default value for
error behaviour.  parse_option() should clean the alternative options and
should not change default value taken from the superblock.

Signed-off-by: Vasily Averin <vvs@sw.ru>
Acked-by: Kirill Korotaev <dev@openvz.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ext3: errors behaviour fix
Dmitry Mishin [Wed, 11 Oct 2006 08:21:49 +0000 (01:21 -0700)]
[PATCH] ext3: errors behaviour fix

Current error behaviour for ext2 and ext3 filesystems does not fully
correspond to the documentation and should be fixed.

According to man 8 mount, ext2 and ext3 file systems allow to set one of 3
different on-errors behaviours:

  ---- start of quote man 8 mount ----

  errors=continue / errors=remount-ro / errors=panic

    Define the behaviour when an error is encountered.  (Either ignore
    errors and just mark the file system erroneous and continue, or remount
    the file system read-only, or panic and halt the system.) The default is
    set in the filesystem superblock, and can be changed using tune2fs(8).

  ---- end of quote ----

However EXT3_ERRORS_CONTINUE is not read from the superblock, and thus
ERRORS_CONT is not saved on the sbi->s_mount_opt.  It leads to the incorrect
handle of errors on ext3.

Then we've checked corresponding code in ext2 and discovered that it is buggy
as well:

- EXT2_ERRORS_CONTINUE is not read from the superblock (the same);

- parse_option() does not clean the alternative values and thus something
  like (ERRORS_CONT|ERRORS_RO) can be set;

- if options are omitted, parse_option() does not set any of these options.

Therefore it is possible to set any combination of these options on the ext2:

- none of them may be set: EXT2_ERRORS_CONTINUE on superblock / empty mount
  options;

- any of them may be set using mount options;

- 2 any options may be set: by using EXT2_ERRORS_RO/EXT2_ERRORS_PANIC on the
  superblock and other value in mount options;

- and finally all three options may be set by adding third option in remount.

Currently ext2 uses these values only in ext2_error() and it is not leading to
any noticeable troubles.  However somebody may be discouraged when he will try
to workaround EXT2_ERRORS_PANIC on the superblock by using errors=continue in
mount options.

This patch:

EXT3_ERRORS_CONTINUE should be taken from the superblock as default value for
error behaviour.

Signed-off-by: Dmitry Mishin <dim@openvz.org>
Acked-by: Vasily Averin <vvs@sw.ru>
Acked-by: Kirill Korotaev <dev@openvz.org>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] fix Module taint flags listing in Oops/panic
Florin Malita [Wed, 11 Oct 2006 08:21:48 +0000 (01:21 -0700)]
[PATCH] fix Module taint flags listing in Oops/panic

Module taint flags listing in Oops/panic has a couple of issues:

* taint_flags() doesn't null-terminate the buffer after printing the flags

* per-module taints are only set if the kernel is not already tainted
  (with that particular flag) => only the first offending module gets its
  taint info correctly updated

Some additional changes:

* 'license_gplok' is no longer needed - equivalent to !(taints &
  TAINT_PROPRIETARY_MODULE) - so we can drop it from struct module *
  exporting module taint info via /proc/module:

pwc 88576 0 - Live 0xf8c32000
evilmod 6784 1 pwc, Live 0xf8bbf000 (PF)

Signed-off-by: Florin Malita <fmalita@gmail.com>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ide-generic: jmicron fix
Alan Cox [Wed, 11 Oct 2006 08:21:47 +0000 (01:21 -0700)]
[PATCH] ide-generic: jmicron fix

Some people find their Jmicron pata port reports its disabled even
though it has devices on it and was boot probed. Fix this

(Candidate for 2.6.18.*, less so for 2.6.19 as we've got a proper
jmicron driver on the merge for that to replace ide-generic support)

Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] grow_buffers() infinite loop fix
Andrew Morton [Wed, 11 Oct 2006 08:21:46 +0000 (01:21 -0700)]
[PATCH] grow_buffers() infinite loop fix

If grow_buffers() is for some reason passed a block number which wants to lie
outside the maximum-addressable pagecache range (PAGE_SIZE * 4G bytes) then it
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: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Add carta_random32() library routine
Stephane Eranian [Wed, 11 Oct 2006 08:21:45 +0000 (01:21 -0700)]
[PATCH] Add carta_random32() library routine

This is a follow-up patch based on the review for perfmon2.  This patch
adds the carta_random32() library routine + carta_random32.h header file.

This is fast, simple, and efficient pseudo number generator algorithm.  We
use it in perfmon2 to randomize the sampling periods.  In this context, we
do not need any fancy randomizer.

Signed-off-by: stephane eranian <eranian@hpl.hp.com>
Cc: David Mosberger <david.mosberger@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] epoll_pwait()
Davide Libenzi [Wed, 11 Oct 2006 08:21:44 +0000 (01:21 -0700)]
[PATCH] epoll_pwait()

Implement the epoll_pwait system call, that extend the event wait mechanism
with the same logic ppoll and pselect do.  The definition of epoll_pwait
is:

int epoll_pwait(int epfd, struct epoll_event *events, int maxevents,
                 int timeout, const sigset_t *sigmask, size_t sigsetsize);

The difference between the vanilla epoll_wait and epoll_pwait is that the
latter allows the caller to specify a signal mask to be set while waiting
for events.  Hence epoll_pwait will wait until either one monitored event,
or an unmasked signal happen.  If sigmask is NULL, the epoll_pwait system
call will act exactly like epoll_wait.  For the POSIX definition of
pselect, information is available here:

http://www.opengroup.org/onlinepubs/009695399/functions/select.html

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Cc: Ulrich Drepper <drepper@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Add CONFIG_HEADERS_CHECK option to automatically run 'make headers_check'
David Woodhouse [Wed, 11 Oct 2006 08:21:43 +0000 (01:21 -0700)]
[PATCH] Add CONFIG_HEADERS_CHECK option to automatically run 'make headers_check'

In order to encourage people to notice when they break the exported
headers, add a config option which automatically runs the sanity checks
when building vmlinux.  That way, those who use allyesconfig will notice
failures.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: allow finer tuning for host VMSPLIT setting
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:42 +0000 (01:21 -0700)]
[PATCH] uml: allow finer tuning for host VMSPLIT setting

Now that various memory splits are enabled, add a config option allowing the
user to compile UML for its need - HOST_2G_2G allowed to choose either 3G/1G
or 2G/2G, and enabling it reduced the usable virtual memory.

Detecting this at run time should be implemented in the future, but we must
make the stop-gap measure work well enough (this is valid in _many_ cases).

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: deprecate CONFIG_MODE_TT
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:41 +0000 (01:21 -0700)]
[PATCH] uml: deprecate CONFIG_MODE_TT

Deprecate TT mode in Kconfig so that users won't select it, update the
MODE_SKAS description (it was largely obsolete and misleadin) and btw describe
advantages for high memory usage with CONFIG_STATIC_LINK.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: remove duplicate export
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:40 +0000 (01:21 -0700)]
[PATCH] uml: remove duplicate export

The export is together with the definition, in arch/x86_64/lib/csum-partial.c,
which is compiled in by arch/um/sys-x86_64/Makefile.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: asm offsets duplication removal
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:40 +0000 (01:21 -0700)]
[PATCH] uml: asm offsets duplication removal

Unify macros common to x86 and x86_64 kernel-offsets.h files.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: allow using again x86/x86_64 crypto code
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:38 +0000 (01:21 -0700)]
[PATCH] uml: allow using again x86/x86_64 crypto code

Enable compilation of x86_64 crypto code;, and add the needed constant to make
the code compile again (that macro was added to i386 asm-offsets between
2.6.17 and 2.6.18, in 6c2bb98bc33ae33c7a33a133a4cd5a06395fece5).

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: declare in Kconfig our partial LOCKDEP support
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:37 +0000 (01:21 -0700)]
[PATCH] uml: declare in Kconfig our partial LOCKDEP support

Declare UML partial support for LOCKDEP - however IRQFLAGS tracing requires
some coding which nobody did yet, so we cannot run full lockdep on UML.  Grep
for CONFIG_TRACE_IRQFLAGS on i386 code to find their implementation.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: fix uname under setarch i386
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:36 +0000 (01:21 -0700)]
[PATCH] uml: fix uname under setarch i386

On a 64bit Uml, if run under "setarch i386" (which a user did), uname()
currently returns the obtained i686 as machine - fix that.  Btw, I'm quite
surprised that under setarch i386 a 64-bit binary can run.

Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: fix processor selection to exclude unsupported processors and features
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:35 +0000 (01:21 -0700)]
[PATCH] uml: fix processor selection to exclude unsupported processors and features

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 do
   not have.

*) 3Dnow support for memcpy() et al. does not compile currently and fixing this
   is not trivial, so simply disable it; with this change, if one selects MK7
   UML compiles (while it did not).

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: make UML_SETJMP always safe
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:34 +0000 (01:21 -0700)]
[PATCH] uml: make UML_SETJMP always safe

If enable is moved by GCC in a register its value may not be preserved after
coming back there with longjmp().  So, mark it as volatile to prevent this;
this is suggested (it seems) in info gcc, when it talks about -Wuninitialized.
 I re-read this and it seems to say something different, but I still believe
this may be needed.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: make TT mode compile after setjmp-related changes
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:34 +0000 (01:21 -0700)]
[PATCH] uml: make TT mode compile after setjmp-related changes

Make TT mode compile after the introduction of klibc's implementation of
setjmp.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: readd forgot prototype
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:33 +0000 (01:21 -0700)]
[PATCH] uml: readd forgot prototype

This was forgot in a previous patch so UML does not compile with TT mode
enabled.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: correct removal of pte_mkexec
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:32 +0000 (01:21 -0700)]
[PATCH] uml: correct removal of pte_mkexec

Correct commit 5906e4171ad61ce68de95e51b773146707671f80 - this makes more
sense: we turn pte_mkexec + pte_wrprotect to pte_mkread.  However, due to a
bug in pte_mkread, it does the exact same thing as pte_mkwrite, so this patch
improves the code but does not change anything in practice.  The pte_mkread
bug is fixed separately, as it may have big impact.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: revert wrong patch
Paolo 'Blaisorblade' Giarrusso [Wed, 11 Oct 2006 08:21:31 +0000 (01:21 -0700)]
[PATCH] uml: revert wrong patch

Andi Kleen pointed out that -mcmodel=kernel does not make sense for userspace
code and would stop everything from working, and pointed out the correct fix
for the original bug (not easy to do for me).

Reverts part of commit 06837504de7b4883e92af207dbbab4310d0db0ed.

Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] mm: locks_freed fix
Nick Piggin [Wed, 11 Oct 2006 08:21:30 +0000 (01:21 -0700)]
[PATCH] mm: locks_freed fix

Move the lock debug checks below the page reserved checks.  Also, having
debug_check_no_locks_freed in kernel_map_pages is wrong.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>