]> git.hungrycats.org Git - linux/log
linux
22 years ago[PATCH] signal handling race fixes: sparc and sparc64
David S. Miller [Wed, 25 Aug 2004 11:01:41 +0000 (04:01 -0700)]
[PATCH] signal handling race fixes: sparc and sparc64

Ok, here are the sparc64 and sparc32 versions.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] signal-race-fixes: x86-64 support
Andi Kleen [Wed, 25 Aug 2004 11:01:29 +0000 (04:01 -0700)]
[PATCH] signal-race-fixes: x86-64 support

  Add the signal race changes to x86-64 to make it compile again.

  Didn't merge the more pointless changes from i386.

  Also remove the special SA_ONESHOT handling, doesn't seem to be needed
  anymore.

From: Mikael Pettersson <mikpe@csd.uu.se>

  The signal-race-fixes patch in 2.6.8-rc2-mm1 appears to have broken
  x86-64's ia32 emulation.

  When forcing a SIGSEGV the old code updated "*ka", where ka was a pointer
  to current's k_sigaction for SIGSEGV.  Now "ka_copy" points to a copy of
  that structure, so assigning "*ka_copy" doesn't do what we want.  Instead do
  the assignment via current->...  just like the normal signal delivery code
  does.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc signal handling fixes
Mikael Pettersson [Wed, 25 Aug 2004 11:01:20 +0000 (04:01 -0700)]
[PATCH] ppc signal handling fixes

2.6.8-rc2-mm1 reintroduced the signal-race-fixes patch for i386, x86_64,
s390, and ia64, breaking all other archs.

The patch below updates ppc, following the pattern of i386.  Compiled &
runtime tested.  No observable breakage.

Signed-off-by: Mikael Pettersson <mikpe@csd.uu.se>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] signal-race fixes for s390
Martin Schwidefsky [Wed, 25 Aug 2004 11:01:08 +0000 (04:01 -0700)]
[PATCH] signal-race fixes for s390

  Update s30 for the signal race fix

From: Mikael Pettersson <mikpe@csd.uu.se>

  The signal-race-fixes patch in 2.6.8-rc2-mm1 appears to be a bit broken on
  s390.

  When forcing a SIGSEGV the old code updated "*ka", where ka was a pointer
  to current's k_sigaction for SIGSEGV.  Now "ka_copy" points to a copy of
  that structure, so assigning "*ka_copy" doesn't do what we want.  Instead do
  the assignment via current->...  just like i386 and x86_64 do.

  Furthermore, the SA_ONESHOT handling wasn't deleted.  That is now handled
  by generic code in the kernel.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] signal handling race fix
Corey Minyard [Wed, 25 Aug 2004 11:00:58 +0000 (04:00 -0700)]
[PATCH] signal handling race fix

The problem:

  In arch/i386/signal.c, in the do_signal() function, it calls
  get_signal_to_deliver() which returns the signal number to deliver (along
  with siginfo).  get_signal_to_deliver() grabs and releases the lock, so
  the signal handler lock is not held in do_signal().  Then the do_signal()
  calls handle_signal(), which uses the signal number to extract the
  sa_handler, etc.

  Since no lock is held, it seems like another thread with the same
  signal handler set can come in and call sigaction(), it can change
  sa_handler between the call to get_signal_to_deliver() and fetching the
  value of sa_handler.  If the sigaction() call set it to SIG_IGN, SIG_DFL,
  or some other fundamental change, that bad things can happen.

The patch:

  You have to get the sigaction information that will be delivered while
  holding sighand->siglock in get_signal_to_deliver().

  In 2.4, it can be fixed per-arch and requires no change to the
  arch-independent code because the arch fetches the signal with
  dequeue_signal() and does all the checking.

The test app:

  The program below has three threads that share signal handlers.  Thread
  1 changes the signal handler for a signal from a handler to SIG_IGN and
  back.  Thread 0 sends signals to thread 3, which just receives them.
  What I believe is happening is that thread 1 changes the signal handler
  in the process of thread 3 receiving the signal, between the time that
  thread 3 fetches the signal info using get_signal_to_deliver() and
  actually delivers the signal with handle_signal().

  Although the program is obvously an extreme case, it seems like any
  time you set the handler value of a signal to SIG_IGN or SIG_DFL, you can
  have this happen.  Changing signal attributes might also cause problems,
  although I am not so sure about that.

  (akpm: this test app segv'd on SMP within milliseconds for me)

#include <signal.h>
#include <stdio.h>
#include <sched.h>

char stack1[16384];
char stack2[16384];

void sighnd(int sig)
{
}

int child1(void *data)
{
struct sigaction act;

sigemptyset(&act.sa_mask);
act.sa_flags = 0;
for (;;) {
act.sa_handler = sighnd;
sigaction(45, &act, NULL);
act.sa_handler = SIG_IGN;
sigaction(45, &act, NULL);
}
}

int child2(void *data)
{
for (;;) {
sleep(100);
}
}

int main(int argc, char *argv[])
{
int pid1, pid2;

signal(45, SIG_IGN);
pid2 = clone(child2, stack2 + sizeof(stack2) - 8,
CLONE_SIGHAND | CLONE_VM, NULL);
pid1 = clone(child1, stack1 + sizeof(stack2) - 8,
CLONE_SIGHAND | CLONE_VM, NULL);

for (;;) {
kill(pid2, 45);
}
}

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years agoMerge common signal handling fault handling in generic code.
Linus Torvalds [Wed, 25 Aug 2004 10:37:34 +0000 (03:37 -0700)]
Merge common signal handling fault handling in generic code.

22 years agovt: don't bother doing UTF translation in control states.
Linus Torvalds [Wed, 25 Aug 2004 10:28:57 +0000 (03:28 -0700)]
vt: don't bother doing UTF translation in control states.

And don't accept UTF translations as the start of a control
state either.

22 years ago[PATCH] mda dependency
Alexander Viro [Wed, 25 Aug 2004 09:11:20 +0000 (02:11 -0700)]
[PATCH] mda dependency

MDA is ISA-only ;-)

22 years ago[PATCH] usb alignment fixes
Alexander Viro [Wed, 25 Aug 2004 09:11:09 +0000 (02:11 -0700)]
[PATCH] usb alignment fixes

Several places did le16_to_cpup() on misaligned address, which blows on
any little-endian platform that doesn't like misaligned reads.

22 years ago[PATCH] warning fix in usb/gadget/inode.c
Alexander Viro [Wed, 25 Aug 2004 09:10:57 +0000 (02:10 -0700)]
[PATCH] warning fix in usb/gadget/inode.c

wrong type of return value.

22 years ago[PATCH] signed char bugs in ixj
Alexander Viro [Wed, 25 Aug 2004 09:10:46 +0000 (02:10 -0700)]
[PATCH] signed char bugs in ixj

Fixed assumption that char is always unsigned

22 years ago[PATCH] killed check_region() in ixj
Alexander Viro [Wed, 25 Aug 2004 09:10:34 +0000 (02:10 -0700)]
[PATCH] killed check_region() in ixj

Killed check_region(), fixed an old bug in ISA case (we checked the wrong
region before claiming the right one - dumb typo back in 2.4.early)

22 years ago[PATCH] annotation of xfs sendfile
Alexander Viro [Wed, 25 Aug 2004 09:10:22 +0000 (02:10 -0700)]
[PATCH] annotation of xfs sendfile

->sendfile() takes kernel pointer, not userland one.

22 years ago[PATCH] annotation of ki_buf
Alexander Viro [Wed, 25 Aug 2004 09:10:10 +0000 (02:10 -0700)]
[PATCH] annotation of ki_buf

->ki_buf is always a userland pointer.

22 years ago[PATCH] removed bogus casts of SPIN_LOCK_UNLOCKED
Alexander Viro [Wed, 25 Aug 2004 09:09:58 +0000 (02:09 -0700)]
[PATCH] removed bogus casts of SPIN_LOCK_UNLOCKED

22 years agoMerge bk://kernel.bkbits.net/gregkh/linux/driver-2.6
Linus Torvalds [Wed, 25 Aug 2004 08:52:57 +0000 (01:52 -0700)]
Merge bk://kernel.bkbits.net/gregkh/linux/driver-2.6
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years agoMerge bk://bk.arm.linux.org.uk/linux-2.6-pcmcia
Linus Torvalds [Wed, 25 Aug 2004 08:50:08 +0000 (01:50 -0700)]
Merge bk://bk.arm.linux.org.uk/linux-2.6-pcmcia
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years agoMerge bk://bk.arm.linux.org.uk/linux-2.6-rmk
Linus Torvalds [Wed, 25 Aug 2004 08:47:07 +0000 (01:47 -0700)]
Merge bk://bk.arm.linux.org.uk/linux-2.6-rmk
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years ago[ADFS] Fix sparse signed bitfield warning
Russell King [Wed, 25 Aug 2004 23:38:53 +0000 (00:38 +0100)]
[ADFS] Fix sparse signed bitfield warning

22 years ago[ARM] Fix some sparse complaints:
Russell King [Wed, 25 Aug 2004 21:13:02 +0000 (22:13 +0100)]
[ARM] Fix some sparse complaints:

Pointers are NULL not 0.
Remove obviously unnecessary iBCS2 shm stuff... we're ARM after all.

22 years ago[ARM PATCH] 2040/1: Increase ARM HARDIRQ_BITS to 9, version 2
Tony Lindgren [Wed, 25 Aug 2004 18:43:04 +0000 (19:43 +0100)]
[ARM PATCH] 2040/1: Increase ARM HARDIRQ_BITS to 9, version 2

Patch from Tony Lindgren

This is an updated version of patch 2004/1 to optimize for immediate constant

22 years ago[ARM PATCH] 2047/1: disable NWFPE_XP on big endian
Lennert Buytenhek [Wed, 25 Aug 2004 18:39:16 +0000 (19:39 +0100)]
[ARM PATCH] 2047/1: disable NWFPE_XP on big endian

Patch from Lennert Buytenhek

Hi,

gcc doesn't understand 80-bit floating point on the ARM currently,
according to the kernel's Kconfig docs, but it would seem that the
current extended double emulation code is broken for big endian
platforms.

So, this patch disables NWFPE_XP on big endian architectures, until
someone comes round and fixes it.

cheers,
Lennert

22 years ago[ARM PATCH] 2046/1: fix nwfpe for double arithmetic on big-endian platforms
Lennert Buytenhek [Wed, 25 Aug 2004 18:35:00 +0000 (19:35 +0100)]
[ARM PATCH] 2046/1: fix nwfpe for double arithmetic on big-endian platforms

Patch from Lennert Buytenhek

Hi,

I need the patch below (against 2.6.8-rc1-ds1) to make nwfpe properly
emulate arithmetic with doubles on a big endian ARM platform.

From reading the mailing list archives and from helpful comments I've
received from people on this list, I gather that this has come up in
the past, but it appears that Russell King was never really convinced
as to why this patch is needed.  I think I understand what's going on,
and will try to explain.

On little endian ARM, the double value 1.0 looks like this when stored
in memory in FPA word ordering:
bytes: 0x00 0x00 0xf0 0x3f 0x00 0x00 0x00 0x00
u32s:  0x3ff00000 0x00000000
u64:   0x000000003ff00000

On big endian, it looks like this:
bytes: 0x3f 0xf0 0x00 0x00 0x00 0x00 0x00 0x00
u32s:  0x3ff00000 0x00000000
u64:   0x3ff0000000000000

It appears to be this way because once upon a time, somebody decided
that the sub-words of a double will use native endian word ordering
within themselves, but the two separate words will always be stored
with the most significant one first.  God knows why they did it this
way, but they did.

Anyway.  The key observation is that nwfpe internally stores double
values in the type 'float64', which is basically just a typedef for
unsigned long long.  It never accesses 'float64's on the byte level
by casting pointers around or anything like that, it just uses direct
u64 arithmetic primitives (add, shift, or, and) for float64
manipulations and that's it.

So.  For little endian platforms, 1.0 looks like:
0x00 0x00 0xf0 0x3f 0x00 0x00 0x00 0x00
But since nwfpe treats it as a u64, it wants it to look like:
0x00 0x00 0x00 0x00 0x00 0x00 0xf0 0x3f
So, that's why the current code swaps the words around when getting
doubles from userspace and putting them back (see fpa11_cpdt.c,
loadDouble and storeDouble.)

On big endian, 1.0 looks like:
0x3f 0xf0 0x00 0x00 0x00 0x00 0x00 0x00
Since nwfpe treats it as a u64, it wants it to look like:
0x3f 0xf0 0x00 0x00 0x00 0x00 0x00 0x00
Hey!  That's exactly the same.  So in this case, it shouldn't be
swapping the halves around.  However, it currently does that swapping
unconditionally, and that's why floating point emulation messes up.

This is how I understand things -- hope it makes sense to other people
too.

cheers,
Lennert

22 years ago[ARM PATCH] 2044/1: S3C2410 - missing IRQ_TICK from RTC resources
Ben Dooks [Wed, 25 Aug 2004 18:18:55 +0000 (19:18 +0100)]
[ARM PATCH] 2044/1: S3C2410 - missing IRQ_TICK from RTC resources

Patch from Ben Dooks

Fixes missing IRQ_TICK from RTC resources

22 years ago[ARM PATCH] 2043/1: S3C2410 update to registered devices
Ben Dooks [Wed, 25 Aug 2004 18:14:27 +0000 (19:14 +0100)]
[ARM PATCH] 2043/1: S3C2410 update to registered devices

Patch from Ben Dooks

Updated all arch/arm/mach-s3c2410/mach-XXX.c files to
register default set of devices

Added new board struct to keep this sort of info, as it
isn't possible to register platform_devices until after
the init_io functions have been called.

22 years ago[ARM PATCH] 2042/1: S3C2410 - Clock fixes, added watchdog clock
Ben Dooks [Wed, 25 Aug 2004 18:09:50 +0000 (19:09 +0100)]
[ARM PATCH] 2042/1: S3C2410 - Clock fixes, added watchdog clock

Patch from Ben Dooks

Added clock definition for watchdog, and fixed it so
that clocks that cannot be enabled/disabled will be
left alone.

Fixed typo in naming of clocks when registering

22 years ago[ARM] Move DMA mask-based bounce detection to dmabounce code.
Russell King [Wed, 25 Aug 2004 15:29:22 +0000 (16:29 +0100)]
[ARM] Move DMA mask-based bounce detection to dmabounce code.

22 years ago[ARM] Correct dma_to_virt()/virt_to_dma() return types.
Russell King [Wed, 25 Aug 2004 14:59:55 +0000 (15:59 +0100)]
[ARM] Correct dma_to_virt()/virt_to_dma() return types.

22 years ago[PATCH] PCI Hotplug: create pci_remove_bus()
John Rose [Wed, 25 Aug 2004 06:23:54 +0000 (23:23 -0700)]
[PATCH] PCI Hotplug: create pci_remove_bus()

The following patch implements a pci_remove_bus() that can be used by
hotplug drivers for the removal of root buses.  It also defines a
release function that frees the device struct for pci_bus->bridge when a
root bus class device is unregistered.

Signed-off-by: John Rose <johnrose@austin.ibm.com>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
22 years ago[PATCH] pci-driver: function documentation fix
François Romieu [Wed, 25 Aug 2004 06:23:15 +0000 (23:23 -0700)]
[PATCH] pci-driver: function documentation fix

The returned value can not be null. Yet its description suggests differently.

From: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
22 years ago[PATCH] I2C: update kernel credits/maintainers
Jean Delvare [Wed, 25 Aug 2004 06:22:38 +0000 (23:22 -0700)]
[PATCH] I2C: update kernel credits/maintainers

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
22 years ago[PATCH] I2C: rename in0_ref to cpu0_vid
Jean Delvare [Wed, 25 Aug 2004 06:21:59 +0000 (23:21 -0700)]
[PATCH] I2C: rename in0_ref to cpu0_vid

This patch changes all the i2c chip drivers and documentation to use the
name "cpu0_vid" instead of "in0_ref". The name "in0_ref" was an error in
the first place as motherboard manufacturers may fail to follow the chip
manufacturer's recommendation about which "in" channel to use for VCore
monitoring.

The new name leaves room for chips able to monitor more than 1 vid
value, such as the LM93 and, to a lesser extent, the PC87360 family (all
by National Semiconductor). These chips are typically designed for
dual-CPU motherboards.

This breaks the interface (obviously) so libsensors has been updated to
support both names.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
22 years ago[PATCH] I2C: keywest class
Jean Delvare [Wed, 25 Aug 2004 06:21:22 +0000 (23:21 -0700)]
[PATCH] I2C: keywest class

This is needed for iBook2 owners to be able to use their ADM1030
hardware monitoring chip. Successfully tested by one user.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
22 years agoMerge kroah.com:/home/greg/linux/BK/bleed-2.6
Greg Kroah-Hartman [Wed, 25 Aug 2004 05:45:48 +0000 (22:45 -0700)]
Merge kroah.com:/home/greg/linux/BK/bleed-2.6
into kroah.com:/home/greg/linux/BK/driver-2.6

22 years agoMerge bk://linux-dj.bkbits.net/cpufreq
Linus Torvalds [Wed, 25 Aug 2004 05:43:29 +0000 (22:43 -0700)]
Merge bk://linux-dj.bkbits.net/cpufreq
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years agokobject: convert struct kobject use kref.
Greg Kroah-Hartman [Wed, 25 Aug 2004 05:30:41 +0000 (22:30 -0700)]
kobject: convert struct kobject use kref.

Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
22 years ago[PATCH] MTD: Additional JEDEC device types
Ian Campbell [Wed, 25 Aug 2004 03:46:37 +0000 (20:46 -0700)]
[PATCH] MTD: Additional JEDEC device types

Add support for a couple of BIOS ROM devices.

The patch has been committed to the MTD CVS tree and adds entries to
jedec_probe.c for AMD AM29F002T, Hyundai HY29F002T and Macronix
MX29F002T parts.

This version is slightly updated from the previous once since I
accidentally added MANUFACTURER_MACRONIX when it already existed.  I
also moved the new definitions to go along with the alphabetical by
manufacturer layout of the file.

Signed-off-by: Ian Campbell <icampbell@arcom.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[CPUFREQ] Remove fsb argument from longhauls calc_speed()
Dave Jones [Tue, 24 Aug 2004 22:15:27 +0000 (23:15 +0100)]
[CPUFREQ] Remove fsb argument from longhauls calc_speed()
It's being passed a global everywhere, so it may as well directly reference it.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] Fix silly typo that broke the compile.
Dave Jones [Tue, 24 Aug 2004 20:17:29 +0000 (21:17 +0100)]
[CPUFREQ] Fix silly typo that broke the compile.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] Work around some broken userspace daemons.
Dave Jones [Tue, 24 Aug 2004 19:49:29 +0000 (20:49 +0100)]
[CPUFREQ] Work around some broken userspace daemons.
Some daemons try to set the speed to the same speed we're currently
running at. Detect that, and bail out early before we fiddle
with registers and such.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] Merge on-demand cpufreq policy governor.
Dave Jones [Tue, 24 Aug 2004 19:27:53 +0000 (20:27 +0100)]
[CPUFREQ] Merge on-demand cpufreq policy governor.

From: "Pallipadi, Venkatesh" <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] Fix reset-to-max-speed on unloading of longhaul driver.
Dave Jones [Tue, 24 Aug 2004 19:21:37 +0000 (20:21 +0100)]
[CPUFREQ] Fix reset-to-max-speed on unloading of longhaul driver.

From: Sven Traenkle

The second part adresses a problem of resetting the max cpu-freq when
unloading the driver. This didn't work for my cpu and I doubt it does
for other. There is no need to pass the computed index of the max.
multiplier in the clock_ratio[] table to longhault_table[] cause the
longhaul_setstate function works with the clock_ratio[] index.

Changing the while loop to a for loop with upper limit isn't actually
necessary as long as the driver is bug free, but thats IMHO not yet the
case, so I suggest this change in order to not loop endlessly or read
beyond the limits of the clock_ratio array.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] Samuel2 can use eblcr to determine FSB.
Dave Jones [Tue, 24 Aug 2004 19:17:48 +0000 (20:17 +0100)]
[CPUFREQ] Samuel2 can use eblcr to determine FSB.

From: Sven Traenkle.

here's a patch that solves some issues I have with the longhaul cpufreq
driver on my epia 6000CL/Via EDEN (actually reporting as CentaurHauls,
family 6, model 7, VIA Samuel 2). The driver tries to compute the fsb speed
while it could actually use the fixed values (as it does for model == 6).
I got this change from the via forum, so no credits to me.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] fix up random CodingStyle /whitespace regressions.
Dave Jones [Tue, 24 Aug 2004 19:11:49 +0000 (20:11 +0100)]
[CPUFREQ] fix up random CodingStyle /whitespace regressions.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] Further cleanups to longhaul driver using defines.
Dave Jones [Tue, 24 Aug 2004 19:08:43 +0000 (20:08 +0100)]
[CPUFREQ] Further cleanups to longhaul driver using defines.
By defining the cpu type at startup, we can make a lot of comparisons
a lot more obvious what they are meaning.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[CPUFREQ] Prettyprint longhaul speeds.
Dave Jones [Tue, 24 Aug 2004 18:48:56 +0000 (19:48 +0100)]
[CPUFREQ] Prettyprint longhaul speeds.
If its >= 1000MHz print it as GHz.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years agoMerge delerium.codemonkey.org.uk:/mnt/data/src/bk/bk-linus
Dave Jones [Tue, 24 Aug 2004 18:15:25 +0000 (19:15 +0100)]
Merge delerium.codemonkey.org.uk:/mnt/data/src/bk/bk-linus
into delerium.codemonkey.org.uk:/mnt/data/src/bk/cpufreq

22 years ago[CPUFREQ] Remove extraneous comment
Dave Jones [Tue, 24 Aug 2004 16:42:53 +0000 (17:42 +0100)]
[CPUFREQ] Remove extraneous comment

Signed-off-by: Dave Jones <davej@redhat.com>
22 years ago[PATCH] ppc64: better handling of H_ENTER failures
Paul Mackerras [Tue, 24 Aug 2004 16:39:21 +0000 (09:39 -0700)]
[PATCH] ppc64: better handling of H_ENTER failures

This changes the hash insertion routines to return an error instead of
calling panic() when HV refuses to insert a HPTE (the hypervisor call to
set up a hashtable PTE is H_ENTER).

The error is now propagated upstream, and either bad_page_fault() is
called (kernel mode) or a SIGBUS signal is forced (user mode).  Some
other panic() cases are also turned into BUG_ON.

Overall, this should provide us with better debugging data if the
problem happens, and avoids errors from userland mapping /dev/mem and
trying to use forbidden IOs (XFree ?) to bring the whole kernel down.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[CPUFREQ] Powersaver also has voltage scaling abilities.
Dave Jones [Tue, 24 Aug 2004 16:38:01 +0000 (17:38 +0100)]
[CPUFREQ] Powersaver also has voltage scaling abilities.
We currently don't do voltage scaling, but we can at least set things up
to prepare for when we do.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years agoUse "insert_resource()" to add the PCI resources to the
Linus Torvalds [Tue, 24 Aug 2004 16:35:36 +0000 (09:35 -0700)]
Use "insert_resource()" to add the PCI resources to the
resource tree.

In contrast to the old "request_resource()", this allows
us to add a resource even when firmware (ACPI) has marked
part of it as being in use.

22 years ago[PATCH] via-rhine: small fixes
Andrew Morton [Tue, 24 Aug 2004 19:35:02 +0000 (15:35 -0400)]
[PATCH] via-rhine: small fixes

From: Roger Luethi <rl@hellgate.ch>

- remove Rhine model names (per Jeff's request)
- remove redundant calls to clear MII cmd
- fill some rhine_private fields earlier

Signed-off-by: Roger Luethi <rl@hellgate.ch>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] via-rhine: de-isolate PHY
Andrew Morton [Tue, 24 Aug 2004 19:34:52 +0000 (15:34 -0400)]
[PATCH] via-rhine: de-isolate PHY

From: Roger Luethi <rl@hellgate.ch>

PHYs may come up isolated.  Make sure we can send data to them.  This code
section needs a clean-up, but I prefer to merge this fix in isolation.

Report and suggested fix by Tam, Ming Dat (Tommy).

Signed-off-by: Roger Luethi <rl@hellgate.ch>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] via-rhine: suspend/resume support
Andrew Morton [Tue, 24 Aug 2004 19:34:42 +0000 (15:34 -0400)]
[PATCH] via-rhine: suspend/resume support

From: Roger Luethi <rl@hellgate.ch>

From: Arkadiusz Miskiewicz

Signed-off-by: Arkadiusz Miskiewicz <arekm@pld-linux.org>
Signed-off-by: Roger Luethi <rl@hellgate.ch>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] netdrv gianfar: fix printk output
Andrew Morton [Tue, 24 Aug 2004 19:34:32 +0000 (15:34 -0400)]
[PATCH] netdrv gianfar: fix printk output

From: Kumar Gala <galak@somerset.sps.mot.com>

Fix usage of printk on the output of mac address.

Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] Typo in drivers/net/dl2k.h
Andrew Morton [Tue, 24 Aug 2004 19:34:20 +0000 (15:34 -0400)]
[PATCH] Typo in drivers/net/dl2k.h

From: Alexander Shatohin <ash@tsi.lv>

Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] 8139too: be sure to progress during rtl8139_rx()
Andrew Morton [Tue, 24 Aug 2004 19:34:10 +0000 (15:34 -0400)]
[PATCH] 8139too: be sure to progress during rtl8139_rx()

From: Francois Romieu <romieu@fr.zoreil.com>

If the Rx buffer gets corrupted or the FIFO hangs in new interesting ways,
this code prevents the driver from looping in ksoftirqd context without
making any progress.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] 8139too: Rx fifo/overflow recovery
Andrew Morton [Tue, 24 Aug 2004 19:34:00 +0000 (15:34 -0400)]
[PATCH] 8139too: Rx fifo/overflow recovery

From: Francois Romieu <romieu@fr.zoreil.com>

This patch allows to update the interrupt status register after an Rx
overflow or a Rx fifo error even when the Rx buffer contains no packet.
The update must be kept in the packet processing loop to prevent an Rx
error storm.  As an interesting behavior, the status of the interrupt
status register must not be read early.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] via-velocity: wrong module name in Kconfig documentation
Andrew Morton [Tue, 24 Aug 2004 19:33:49 +0000 (15:33 -0400)]
[PATCH] via-velocity: wrong module name in Kconfig documentation

From: Francois Romieu <romieu@fr.zoreil.com>

Copy/paste abuse.

Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] drivers/net/wan/cycx_x25.c:189: warning: conflicting types for built-in funct...
Andrew Morton [Tue, 24 Aug 2004 19:33:39 +0000 (15:33 -0400)]
[PATCH] drivers/net/wan/cycx_x25.c:189: warning: conflicting types for built-in function 'log2'

From: Jesper Juhl <juhl-lkml@dif.dk>

To silence the warning in $subject, rename log2 to cycx_log2 in this file
to remove the clash, so there's no doubt that this file uses it's own
defined log2 function.

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] fix net/hamradio/dmascc with gcc 3.4
Andrew Morton [Tue, 24 Aug 2004 19:26:50 +0000 (15:26 -0400)]
[PATCH] fix net/hamradio/dmascc with gcc 3.4

From: Adrian Bunk <bunk@fs.tum.de>

drivers/net/hamradio/dmascc.c: In function `scc_isr':
drivers/net/hamradio/dmascc.c:250: sorry, unimplemented: inlining
failed in call to 'z8530_isr': function body not available
drivers/net/hamradio/dmascc.c:969: sorry, unimplemented: called from
here
drivers/net/hamradio/dmascc.c:250: sorry, unimplemented: inlining
failed in call to 'z8530_isr': function body not available
drivers/net/hamradio/dmascc.c:978: sorry, unimplemented: called from
here

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] sk98lin/skge.c doesn't compile with PROC_FS=n
Andrew Morton [Tue, 24 Aug 2004 19:26:40 +0000 (15:26 -0400)]
[PATCH] sk98lin/skge.c doesn't compile with PROC_FS=n

From: Adrian Bunk <bunk@fs.tum.de>

drivers/net/sk98lin/skge.c: In function `skge_remove_one':
drivers/net/sk98lin/skge.c:5116: warning: implicit declaration of function `remove_proc_entry'
drivers/net/sk98lin/skge.c:5116: `pSkRootDir' undeclared (first use in this function)
drivers/net/sk98lin/skge.c:5116: (Each undeclared identifier is reported only once
drivers/net/sk98lin/skge.c:5116: for each function it appears in.)
drivers/net/sk98lin/skge.c: In function `skge_init':
drivers/net/sk98lin/skge.c:5188: `SK_Root_Dir_entry' undeclared (first use in this function)

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] via-velocity: more inetaddr_notifier fix
Andrew Morton [Tue, 24 Aug 2004 19:26:28 +0000 (15:26 -0400)]
[PATCH] via-velocity: more inetaddr_notifier fix

From: Francois Romieu <romieu@fr.zoreil.com>

There is no guarantee that the event which gets passed is associated to a
via-velocity device, thus preventing to dereference dev->priv as if it
always was a struct velocity_info *.  The via-velocity devices are kept in
a module private list for comparison.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] ixgb_main.c: fix inline compile errors
Andrew Morton [Tue, 24 Aug 2004 19:26:18 +0000 (15:26 -0400)]
[PATCH] ixgb_main.c: fix inline compile errors

From: Adrian Bunk <bunk@fs.tum.de>

drivers/net/ixgb/ixgb_main.c: In function `ixgb_up':
drivers/net/ixgb/ixgb_main.c:86: sorry, unimplemented: inlining failed
in call to 'ixgb_irq_enable': function body not available
drivers/net/ixgb/ixgb_main.c:234: sorry, unimplemented: called from here

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] net/tulip/dmfe.c: gcc-3.5 fixes
Andrew Morton [Tue, 24 Aug 2004 19:26:07 +0000 (15:26 -0400)]
[PATCH] net/tulip/dmfe.c: gcc-3.5 fixes

From: Adrian Bunk <bunk@fs.tum.de>

  CC      drivers/net/tulip/dmfe.o
drivers/net/tulip/dmfe.c: In function `dmfe_rx_packet':
drivers/net/tulip/dmfe.c:323: sorry, unimplemented: inlining failed in
call to 'cal_CRC': function body not available
drivers/net/tulip/dmfe.c:936: sorry, unimplemented: called from here
make[3]: *** [drivers/net/tulip/dmfe.o] Error 1

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] net/rrunner.c: gcc-3.5 fixes
Andrew Morton [Tue, 24 Aug 2004 19:25:57 +0000 (15:25 -0400)]
[PATCH] net/rrunner.c: gcc-3.5 fixes

From: Adrian Bunk <bunk@fs.tum.de>

  CC      drivers/net/rrunner.o
drivers/net/rrunner.c: In function `rr_timer':
drivers/net/rrunner.h:846: sorry, unimplemented: inlining failed in call
to 'rr_raz_tx': function body not available
drivers/net/rrunner.c:1155: sorry, unimplemented: called from here
drivers/net/rrunner.h:847: sorry, unimplemented: inlining failed in call
to 'rr_raz_rx': function body not available
drivers/net/rrunner.c:1156: sorry, unimplemented: called from here
make[2]: *** [drivers/net/rrunner.o] Error 1

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] net/hamachi.c: gcc-3.5 build fixes
Andrew Morton [Tue, 24 Aug 2004 19:25:47 +0000 (15:25 -0400)]
[PATCH] net/hamachi.c: gcc-3.5 build fixes

From: Adrian Bunk <bunk@fs.tum.de>

  CC      drivers/net/hamachi.o
drivers/net/hamachi.c: In function `hamachi_interrupt':
drivers/net/hamachi.c:562: sorry, unimplemented: inlining failed in call
to 'hamachi_rx': function body not available
drivers/net/hamachi.c:1402: sorry, unimplemented: called from here
make[2]: *** [drivers/net/hamachi.o] Error 1

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] net/smc9194.c: fix gcc-3.5 inline compile errors
Andrew Morton [Tue, 24 Aug 2004 19:25:36 +0000 (15:25 -0400)]
[PATCH] net/smc9194.c: fix gcc-3.5 inline compile errors

From: Adrian Bunk <bunk@fs.tum.de>

  CC      drivers/net/smc9194.o
drivers/net/smc9194.c: In function `smc_interrupt':
drivers/net/smc9194.c:278: sorry, unimplemented: inlining failed in call
to 'smc_rcv': function body not available
drivers/net/smc9194.c:1254: sorry, unimplemented: called from here
drivers/net/smc9194.c:283: sorry, unimplemented: inlining failed in call
to 'smc_tx': function body not available
drivers/net/smc9194.c:1258: sorry, unimplemented: called from here
make[2]: *** [drivers/net/smc9194.o] Error 1

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] e1000 inlining fix
Andrew Morton [Tue, 24 Aug 2004 19:25:26 +0000 (15:25 -0400)]
[PATCH] e1000 inlining fix

From: Nick Orlov <bugfixer@list.ru>

e1000 fixes for gcc-3.4.1

Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] e1000 build fix
Andrew Morton [Tue, 24 Aug 2004 19:25:15 +0000 (15:25 -0400)]
[PATCH] e1000 build fix

drivers/net/e1000/e1000_main.c: In function `e1000_up':
drivers/net/e1000/e1000_main.c:136: sorry, unimplemented: inlining failed
in call to 'e1000_irq_enable': function body not available
drivers/net/e1000/e1000_main.c:274: sorry, unimplemented: called from here

Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] sk98lin procfs fix
Andrew Morton [Tue, 24 Aug 2004 19:25:05 +0000 (15:25 -0400)]
[PATCH] sk98lin procfs fix

From: Christoph Hellwig <hch@infradead.org>

sk98line tries to register a procfile with the interfacename of the struct
net_device.  The patch below (ontop of the previous one) makes it work
unless you change the interface name manually, but as Linux explicitly
allows that the interface is fundamentally broken and probably should just
go away.

Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[PATCH] R8169_NAPI help text
Andrew Morton [Tue, 24 Aug 2004 19:24:55 +0000 (15:24 -0400)]
[PATCH] R8169_NAPI help text

From: Adrian Bunk <bunk@fs.tum.de>

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
22 years ago[CPUFREQ] Introduce some defines for the longhaul version, and use them.
Dave Jones [Tue, 24 Aug 2004 16:35:28 +0000 (17:35 +0100)]
[CPUFREQ] Introduce some defines for the longhaul version, and use them.
This makes some of the code quite a bit cleaner, and a lot more
obvious whats going on on which CPUs.

Signed-off-by: Dave Jones <davej@redhat.com>
22 years agoMerge bk://linux-sam.bkbits.net/kbuild
Linus Torvalds [Tue, 24 Aug 2004 12:11:17 +0000 (05:11 -0700)]
Merge bk://linux-sam.bkbits.net/kbuild
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years agokbuild: fix cc-version
Sam Ravnborg [Wed, 25 Aug 2004 00:36:35 +0000 (02:36 +0200)]
kbuild: fix cc-version

cc-version needs to use $(shell to get the gcc version.
Before if gave the following error when building the kernel:

/bin/sh: line 1: [: too many arguments

And all checks for gcc version were broken.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
22 years agoMerge mars.ravnborg.org:/home/sam/bk/linux-2.6
Sam Ravnborg [Wed, 25 Aug 2004 00:24:31 +0000 (02:24 +0200)]
Merge mars.ravnborg.org:/home/sam/bk/linux-2.6
into mars.ravnborg.org:/home/sam/bk/kbuild

22 years agoMerge mars.ravnborg.org:/home/sam/bk/linux-2.6
Sam Ravnborg [Tue, 24 Aug 2004 23:23:46 +0000 (01:23 +0200)]
Merge mars.ravnborg.org:/home/sam/bk/linux-2.6
into mars.ravnborg.org:/home/sam/bk/kbuild

22 years ago[PATCH] kobject: fix kobject_set_name comment.
Dmitry Torokhov [Tue, 24 Aug 2004 09:19:31 +0000 (02:19 -0700)]
[PATCH] kobject: fix kobject_set_name comment.

Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
22 years agoMerge bk://kernel.bkbits.net/davem/net-2.6
Linus Torvalds [Tue, 24 Aug 2004 05:43:49 +0000 (22:43 -0700)]
Merge bk://kernel.bkbits.net/davem/net-2.6
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years agoMerge bk://ppc.bkbits.net/for-linus-ppc
Linus Torvalds [Tue, 24 Aug 2004 05:40:58 +0000 (22:40 -0700)]
Merge bk://ppc.bkbits.net/for-linus-ppc
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years ago[PATCH] GPCMD_SEND_CUE_SHEET missing in scsi_ioctl
Jens Axboe [Tue, 24 Aug 2004 05:33:27 +0000 (22:33 -0700)]
[PATCH] GPCMD_SEND_CUE_SHEET missing in scsi_ioctl

Forgot one command, GPCMD_SEND_CUE_SHEET is also ok for write open.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] /dev/ptmx open() fixes
Alexander Viro [Tue, 24 Aug 2004 05:29:18 +0000 (22:29 -0700)]
[PATCH] /dev/ptmx open() fixes

If tty_open() fails for a normal serial device, we end up doing cleanups
that should only happen for failed open of /dev/ptmx.  The results are
not pretty - devpts et.al.  end up very confused.  That's what gave
problems with ptmx.

This splits ptmx file_operations from the normal case and cleans up both
tty_open() and (new) ptmx_open().  Survived serious beating.

22 years ago[PATCH] Missing free_area_init_node() conversions
William Lee Irwin III [Tue, 24 Aug 2004 05:29:06 +0000 (22:29 -0700)]
[PATCH] Missing free_area_init_node() conversions

Update architectures for the free_area_init_node() API change.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] Undo broken FH conversion that broke nfsroot compile
Trond Myklebust [Tue, 24 Aug 2004 05:28:55 +0000 (22:28 -0700)]
[PATCH] Undo broken FH conversion that broke nfsroot compile

That conversion to nfs_fh_copy() is bogus since we're not copying into
an nfs_fh anyway. Just revert it.

22 years ago[PATCH] Alex DeVries has moved
Adrian Bunk [Tue, 24 Aug 2004 04:46:22 +0000 (21:46 -0700)]
[PATCH] Alex DeVries has moved

The patch below replaces all occurences of two bouncing email addresses of
Alex deVries in the kernel with his current address.

It's already ACK'ed by Alex deVries.

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] describe Intel cache descriptors.
Dave Jones [Tue, 24 Aug 2004 04:46:09 +0000 (21:46 -0700)]
[PATCH] describe Intel cache descriptors.

Describe what the Intel cache descriptors actually mean in comments.  Taken
from 24151827.pdf.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] Fix fs/locks.c init order
Arjan van de Ven [Tue, 24 Aug 2004 04:45:58 +0000 (21:45 -0700)]
[PATCH] Fix fs/locks.c init order

The patch below fixes an interesting oddity we're seeing with fedora core
development (where we recently started using udev heavily); basically right
now filelock_init() is a module_init(), eg runs late.  However that breaks
down because there are earlier /sbin/hotplug callouts, which with udev, do
locking operations.  When that happens the kernel oopses because the slabs
for file locks aren't initialized yet.

Solution: initialize this way early.  It's only a kmem_cache_create after
all, so can happen early.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] remove obsolete zero-paged in Documentation/sysctl/kernel.txt
Olaf Hering [Tue, 24 Aug 2004 04:45:48 +0000 (21:45 -0700)]
[PATCH] remove obsolete zero-paged in Documentation/sysctl/kernel.txt

This entry was removed during 2.5 development.

Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] remove obsolete htab-reclaim in Documentation/sysctl/kernel.txt
Olaf Hering [Tue, 24 Aug 2004 04:45:37 +0000 (21:45 -0700)]
[PATCH] remove obsolete htab-reclaim in Documentation/sysctl/kernel.txt

This entry is long gone, even 2.4 doesnt have it anymore.

Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] inode time update funnies in ncpfs
Christoph Hellwig [Tue, 24 Aug 2004 04:45:25 +0000 (21:45 -0700)]
[PATCH] inode time update funnies in ncpfs

ncfpfs seems to update inode times by hand everywhere instead of using
the proper helpers.  This means:

 - the atime updates in mmap() and read() seems to miss various checks
   upodate_atime or one of the wrappers does.  Also it doesn't mark the
   inode dirty.
 - in write() you update mtime and _a_time instead of ctime as expected,
   also the usual checks and optimizations are missing.

In addition the fops contain some bogus checks like for a refular file (but
the fops are only used of ISREG files) and inode->i_sb although that is
guranteed to be non-zero.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] show Active/Inactive on per-node meminfo
Akinobu Mita [Tue, 24 Aug 2004 04:45:14 +0000 (21:45 -0700)]
[PATCH] show Active/Inactive on per-node meminfo

  The patch below enable to display the size of Active/Inactive pages on
  per-node meminfo (/sys/devices/system/node/node%d/meminfo) like
  /proc/meminfo.

  By a little change to procps, "vmstat -a" can show these statistics about
  particular node.

From: mita akinobu <amgta@yacht.ocn.ne.jp>

  get_zone_counts() is used by max_sane_readahead(), and
  max_sane_readahead() is often called in filemap_nopage().

Signed-off-by: Akinobu Mita <amgta@yacht.ocn.ne.jp>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] Fix bad URL in BSD acct help entry
Tim Schmielau [Tue, 24 Aug 2004 04:45:02 +0000 (21:45 -0700)]
[PATCH] Fix bad URL in BSD acct help entry

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] /dev/random: Remove RNDGETPOOL ioctl
Theodore Y. Ts'o [Tue, 24 Aug 2004 04:44:50 +0000 (21:44 -0700)]
[PATCH] /dev/random: Remove RNDGETPOOL ioctl

Recently, someone has kvetched that RNDGETPOOL is a "security
vulnerability".  Never mind that it is superuser only, and with superuser
privs you could load a nasty kernel module, or read the entropy pool out of
/dev/mem directly, but they are nevertheless still spreading FUD.

In any case, no one is using it (it was there for debugging purposes only),
so we can remove it as dead code.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] /dev/random: Use separate entropy store for /dev/urandom
Theodore Y. Ts'o [Tue, 24 Aug 2004 04:44:39 +0000 (21:44 -0700)]
[PATCH] /dev/random: Use separate entropy store for /dev/urandom

This patch adds a separate pool for use with /dev/urandom.  This prevents a
/dev/urandom read from being able to completely drain the entropy in the
/dev/random pool, and also makes it much more difficult for an attacker to
carry out a state extension attack.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] /dev/random: Add pool name to entropy store
Theodore Y. Ts'o [Tue, 24 Aug 2004 04:44:27 +0000 (21:44 -0700)]
[PATCH] /dev/random: Add pool name to entropy store

This adds a pool name to the entropy_store data structure, which simplifies
the debugging code, and makes the code more generic for adding additional
entropy pools.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] dev/random: Fix latency in rekeying sequence number
Theodore Y. Ts'o [Tue, 24 Aug 2004 04:44:18 +0000 (21:44 -0700)]
[PATCH] dev/random: Fix latency in rekeying sequence number

Based on reports from Ingo's Latency Tracer that the TCP sequence number
rekey code is causing latency problems, I've moved the sequence number
rekey to be done out of a workqueue.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] file_ra_state_init speedup
Andrew Morton [Tue, 24 Aug 2004 04:44:06 +0000 (21:44 -0700)]
[PATCH] file_ra_state_init speedup

Marcelo points out that this function's main caller already memsets the
structure, so avoid doing it again.

Also, an earlier knfsd patch withdrew file_ra_state_init()'s other caller, so
unexport this function.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] Firmware Loader is orphan
Ramón Rey Vicente [Tue, 24 Aug 2004 04:43:55 +0000 (21:43 -0700)]
[PATCH] Firmware Loader is orphan

The author and maintainer of the firmware loader died in May.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ad1816 sound driver web page and email address
Thorsten Knabe [Tue, 24 Aug 2004 04:43:44 +0000 (21:43 -0700)]
[PATCH] ad1816 sound driver web page and email address

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ext3 documentation
Diego Calleja García [Tue, 24 Aug 2004 04:43:32 +0000 (21:43 -0700)]
[PATCH] ext3 documentation

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>