Jan Beulich [Fri, 5 Oct 2012 00:13:24 +0000 (17:13 -0700)]
lib/vsprintf.c: improve standard conformance of sscanf()
Xen's pciback points out a couple of deficiencies with vsscanf()'s
standard conformance:
- Trailing character matching cannot be checked by the caller: With a
format string of "(%x:%x.%x) %n" absence of the closing parenthesis
cannot be checked, as input of "(00:00.0)" doesn't cause the %n to be
evaluated (because of the code not skipping white space before the
trailing %n).
- The parameter corresponding to a trailing %n could get filled even if
there was a matching error: With a format string of "(%x:%x.%x)%n",
input of "(00:00.0]" would still fill the respective variable pointed to
(and hence again make the mismatch non-detectable by the caller).
This patch aims at fixing those, but leaves other non-conforming aspects
of it untouched, among them these possibly relevant ones:
- improper handling of the assignment suppression character '*' (blindly
discarding all succeeding non-white space from the format and input
strings),
- not honoring conversion specifiers for %n, - not recognizing the C99
conversion specifier 't' (recognized by vsprintf()).
lib/spinlock_debug: avoid livelock in do_raw_spin_lock()
The logic in do_raw_spin_lock() attempts to acquire a spinlock by invoking
arch_spin_trylock() in a loop with a delay between each attempt. Now
consider the following situation in a 2 CPU system:
1. CPU-0 continually acquires and releases a spinlock in a
tight loop; it stays in this loop until some condition X
is satisfied. X can only be satisfied by another CPU.
2. CPU-1 tries to acquire the same spinlock, in an attempt
to satisfy the aforementioned condition X. However, it
never sees the unlocked value of the lock because the
debug spinlock code uses trylock instead of just lock;
it checks at all the wrong moments - whenever CPU-0 has
locked the lock.
Now in the absence of debug spinlocks, the architecture specific spinlock
code can correctly allow CPU-1 to wait in a "queue" (e.g., ticket
spinlocks), ensuring that it acquires the lock at some point. However,
with the debug spinlock code, livelock can easily occur due to the use of
try_lock, which obviously cannot put the CPU in that "queue". This
queueing mechanism is implemented in both x86 and ARM spinlock code.
Note that the situation mentioned above is not hypothetical. A real
problem was encountered where CPU-0 was running hrtimer_cancel with
interrupts disabled, and CPU-1 was attempting to run the hrtimer that
CPU-0 was trying to cancel.
Address this by actually attempting arch_spin_lock once it is suspected
that there is a spinlock lockup. If we're in a situation that is
described above, the arch_spin_lock should succeed; otherwise other
timeout mechanisms (e.g., watchdog) should alert the system of a lockup.
Therefore, if there is a genuine system problem and the spinlock can't be
acquired, the end result (irrespective of this change being present) is
the same. If there is a livelock caused by the debug code, this change
will allow the lock to be acquired, depending on the implementation of the
lower level arch specific spinlock code.
genalloc: make it possible to use a custom allocation algorithm
Premit use of another algorithm than the default first-fit one. For
example a custom algorithm could be used to manage alignment requirements.
As I can't predict all the possible requirements/needs for all allocation
uses cases, I add a "free" field 'void *data' to pass any needed
information to the allocation function. For example 'data' could be used
to handle a structure where you store the alignment, the expected memory
bank, the requester device, or any information that could influence the
allocation algorithm.
An usage example may look like this:
struct my_pool_constraints {
int align;
int bank;
...
};
unsigned long my_custom_algo(unsigned long *map, unsigned long size,
unsigned long start, unsigned int nr, void *data)
{
struct my_pool_constraints *constraints = data;
...
deal with allocation contraints
...
return the index in bitmap where perform the allocation
}
Add of best-fit algorithm function:
most of the time best-fit is slower then first-fit but memory fragmentation
is lower. The random buffer allocation/free tests don't show any arithmetic
relation between the allocation time and fragmentation but the
best-fit algorithm
is sometime able to perform the allocation when the first-fit can't.
This new algorithm help to remove static allocations on ESRAM, a small but
fast on-chip RAM of few KB, used for high-performance uses cases like DMA
linked lists, graphic accelerators, encoders/decoders. On the Ux500
(in the ARM tree) we have define 5 ESRAM banks of 128 KB each and use of
static allocations becomes unmaintainable:
cd arch/arm/mach-ux500 && grep -r ESRAM .
./include/mach/db8500-regs.h:/* Base address and bank offsets for ESRAM */
./include/mach/db8500-regs.h:#define U8500_ESRAM_BASE 0x40000000
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK_SIZE 0x00020000
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK0 U8500_ESRAM_BASE
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK1 (U8500_ESRAM_BASE + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK2 (U8500_ESRAM_BANK1 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK3 (U8500_ESRAM_BANK2 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK4 (U8500_ESRAM_BANK3 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_DMA_LCPA_OFFSET 0x10000
./include/mach/db8500-regs.h:#define U8500_DMA_LCPA_BASE
(U8500_ESRAM_BANK0 + U8500_ESRAM_DMA_LCPA_OFFSET)
./include/mach/db8500-regs.h:#define U8500_DMA_LCLA_BASE U8500_ESRAM_BANK4
I want to use genalloc to do dynamic allocations but I need to be able to
fine tune the allocation algorithm. I my case best-fit algorithm give
better results than first-fit, but it will not be true for every use case.
Jan Beulich [Fri, 5 Oct 2012 00:13:17 +0000 (17:13 -0700)]
lib/Kconfig.debug: adjust hard-lockup related Kconfig options
The main option should not appear in the resulting .config when the
dependencies aren't met (i.e. use "depends on" rather than directly
setting the default from the combined dependency values).
The sub-options should depend on the main option rather than a more
generic higher level one.
Alex Elder [Fri, 5 Oct 2012 00:13:16 +0000 (17:13 -0700)]
lib/parser.c: avoid overflow in match_number()
The result of converting an integer value to another signed integer type
that's unable to represent the original value is implementation defined.
(See notes in section 6.3.1.3 of the C standard.)
In match_number(), the result of simple_strtol() (which returns type long)
is assigned to a value of type int.
Instead, handle the result of simple_strtol() in a well-defined way, and
return -ERANGE if the result won't fit in the int variable used to hold
the parsed result.
No current callers pay attention to the particular error value returned,
so this additional return code shouldn't do any harm.
This driver was for the ProGear webpad device which was produced in
2000/2001 and is not available on a market. I no longer have this
hardware so can not even check how Linux works on it.
G.Shark Jeong [Fri, 5 Oct 2012 00:12:55 +0000 (17:12 -0700)]
backlight: add new lm3639 backlight driver
This driver is a general version for LM3639 backlgiht + flash driver chip
of TI.
LM3639:
The LM3639 is a single chip LCD Display Backlight driver + white LED
Camera driver. Programming is done over an I2C compatible interface.
www.ti.com
G.Shark Jeong [Fri, 5 Oct 2012 00:12:52 +0000 (17:12 -0700)]
backlight: add Backlight driver for lm3630 chip
This driver is a general version for LM3630 backlgiht driver chip of TI.
LM3630 :
The LM3630 is a current mode boost converter which supplies the power
and controls the current in two strings of up to 10 LEDs per string.
Programming is done over an I2C compatible interface.
www.ti.com
hongfeng [Fri, 5 Oct 2012 00:12:25 +0000 (17:12 -0700)]
poweroff: fix bug in orderly_poweroff()
orderly_poweroff is trying to poweroff platform in two steps:
step 1: Call user space application to poweroff
step 2: If user space poweroff fail, then do a force power off if force param
is set.
The bug here is, step 1 is always successful with param UMH_NO_WAIT, which obey
the design goal of orderly_poweroff.
We have two choices here:
UMH_WAIT_EXEC which means wait for the exec, but not the process;
UMH_WAIT_PROC which means wait for the process to complete.
we need to trade off the two choices:
If using UMH_WAIT_EXEC, there is potential issue comments by Serge E.
Hallyn: The exec will have started, but may for whatever (very unlikely)
reason fail.
If using UMH_WAIT_PROC, there is potential issue comments by Eric W.
Biederman: If the caller is not running in a kernel thread then we can
easily get into a case where the user space caller will block waiting for
us when we are waiting for the user space caller.
Thanks for their excellent ideas, based on the above discussion, we
finally choose UMH_WAIT_EXEC, which is much more safe, if the user
application really fails, we just complain the application itself, it
seems a better choice here.
Shawn Guo [Fri, 5 Oct 2012 00:12:23 +0000 (17:12 -0700)]
kernel/sys.c: call disable_nonboot_cpus() in kernel_restart()
As kernel_power_off() calls disable_nonboot_cpus(), we may also want to
have kernel_restart() call disable_nonboot_cpus(). Doing so can help
machines that require boot cpu be the last alive cpu during reboot to
survive with kernel restart.
This fixes one reboot issue seen on imx6q (Cortex-A9 Quad). The machine
requires that the restart routine be run on the primary cpu rather than
secondary ones. Otherwise, the secondary core running the restart
routine will fail to come to online after reboot.
Jiri Kosina [Fri, 5 Oct 2012 00:12:22 +0000 (17:12 -0700)]
tile: fix personality bits handling upon exec()
Historically, the top three bytes of personality have been used for
things such as ADDR_NO_RANDOMIZE, which made sense only for specific
architectures.
We now however have a flag there that is general no matter the
architecture (UNAME26); generally we have to be careful to preserve the
personality flags across exec().
This patch fixes tile architecture not to forcefully overwrite
personality flags during exec().
In addition to that, we fix two other things along the way:
- exec_domain switching is fixed -- set_personality() should always
be used instead of directly assigning to current->personality.
- as pointed out by Arnd Bergmann, PER_LINUX_32BIT is not used anywhere
by tile, so let's just drop that in favor of PER_LINUX
Jiri Kosina [Fri, 5 Oct 2012 00:12:20 +0000 (17:12 -0700)]
cross-arch: don't corrupt personality flags upon exec()
Historically, the top three bytes of personality have been used for
things such as ADDR_NO_RANDOMIZE, which made sense only for specific
architectures.
We now however have a flag there that is general no matter the
architecture (UNAME26); generally we have to be careful to preserve the
personality flags across exec().
This patch tries to fix all architectures that forcefully overwrite
personality flags during exec() (ppc32 and s390 have been fixed recently
by commits f9783ec862ea ("[S390] Do not clobber personality flags on
exec") and 59e4c3a2fe9c ("powerpc/32: Don't clobber personality flags on
exec") in a similar way already).
Commit 6afe1a1fe8ff ("PM: Remove legacy PM") removed the initialization
of retval, causing:
arch/frv/kernel/pm.c: In function 'sysctl_pm_do_suspend':
arch/frv/kernel/pm.c:165:5: warning: 'retval' may be used uninitialized in this function [-Wuninitialized]
Remove the variable completely to fix this, and convert to a proper
switch (...) { ... } construct to improve readability.
Joe Mario [Fri, 5 Oct 2012 00:12:15 +0000 (17:12 -0700)]
sections: fix const sections for crc32 table
Fix the const sections for the code generated by crc32 table. There's
no ro version of the cacheline aligned section, so we cannot put in
const data without a conflict Just don't make the crc tables const for
now.
Andi Kleen [Fri, 5 Oct 2012 00:11:27 +0000 (17:11 -0700)]
sections: disable const sections for PA-RISC v2
The PA-RISC tool chain seems to have some problem with correct
read/write attributes on sections. This causes problems when the const
sections are fixed up for other architecture to only contain truly
read-only data.
The argument to __const_udelay is the number of jiffies to wait divided
by 4, but this does not work unless the multiplication does not
overflow, and that is what the build error is designed to prevent. The
intended behavior can be achieved by using mdelay to call udelay
multiple times in a loop.
Sascha Hauer [Fri, 5 Oct 2012 00:11:17 +0000 (17:11 -0700)]
kbuild: make: fix if_changed when command contains backslashes
The call if_changed mechanism does not work when the command contains
backslashes. This basically is an issue with lzo and bzip2 compressed
kernels. The compressed binaries do not contain the uncompressed image
size, so these use size_append to append the size. This results in
backslashes in the executed command. With this if_changed always
detects a change in the command and rebuilds the compressed image even
if nothing has changed.
Fabio Estevam [Fri, 5 Oct 2012 00:11:16 +0000 (17:11 -0700)]
drivers/dma/dmaengine.c: lower the priority of 'failed to get' dma channel message
Do the same as commit a03a202e95fd ("dmaengine: failure to get a
specific DMA channel is not critical") to get rid of the following
messages during kernel boot:
dmaengine_get: failed to get dma1chan0: (-22)
dmaengine_get: failed to get dma1chan1: (-22)
dmaengine_get: failed to get dma1chan2: (-22)
dmaengine_get: failed to get dma1chan3: (-22)
..
David Daney [Fri, 5 Oct 2012 17:45:30 +0000 (10:45 -0700)]
Partially revert a1ce39288e6fbef (UAPI: (Scripted) Convert #include "..." to #include <path/...> in kernel system headers)
Syntax errors were introduced into include/linux/libfdt.h by the
offending commit, revert the changes made to this file. The kernel
again compiles, thus restoring harmony and balance to the universe.
Dmitry Monakhov [Fri, 5 Oct 2012 15:31:55 +0000 (11:31 -0400)]
ext4: fix ext4_flush_completed_IO wait semantics
BUG #1) All places where we call ext4_flush_completed_IO are broken
because buffered io and DIO/AIO goes through three stages
1) submitted io,
2) completed io (in i_completed_io_list) conversion pended
3) finished io (conversion done)
And by calling ext4_flush_completed_IO we will flush only
requests which were in (2) stage, which is wrong because:
1) punch_hole and truncate _must_ wait for all outstanding unwritten io
regardless to it's state.
2) fsync and nolock_dio_read should also wait because there is
a time window between end_page_writeback() and ext4_add_complete_io()
As result integrity fsync is broken in case of buffered write
to fallocated region:
fsync blkdev_completion
->filemap_write_and_wait_range
->ext4_end_bio
->end_page_writeback
<-- filemap_write_and_wait_range return
->ext4_flush_completed_IO
sees empty i_completed_io_list but pended
conversion still exist
->ext4_add_complete_io
BUG #2) Race window becomes wider due to the 'ext4: completed_io
locking cleanup V4' patch series
This patch make following changes:
1) ext4_flush_completed_io() now first try to flush completed io and when
wait for any outstanding unwritten io via ext4_unwritten_wait()
2) Rename function to more appropriate name.
3) Assert that all callers of ext4_flush_unwritten_io should hold i_mutex to
prevent endless wait
Merge branch 'staging/for_v3.7' into v4l_for_linus
* staging/for_v3.7: (2891 commits)
em28xx: regression fix: use DRX-K sync firmware requests on em28xx
drxk: allow loading firmware synchrousnously
em28xx: Make all em28xx extensions to be initialized asynchronously
[media] tda18271: properly report read errors in tda18271_get_id
[media] tda18271: delay IR & RF calibration until init() if delay_cal is set
[media] MAINTAINERS: add Michael Krufky as tda827x maintainer
[media] MAINTAINERS: add Michael Krufky as tda8290 maintainer
[media] MAINTAINERS: add Michael Krufky as cxusb maintainer
[media] MAINTAINERS: add Michael Krufky as lg2160 maintainer
[media] MAINTAINERS: add Michael Krufky as lgdt3305 maintainer
[media] MAINTAINERS: add Michael Krufky as mxl111sf maintainer
[media] MAINTAINERS: add Michael Krufky as mxl5007t maintainer
[media] MAINTAINERS: add Michael Krufky as tda18271 maintainer
[media] s5p-tv: Report only multi-plane capabilities in vidioc_querycap
[media] s5p-mfc: Fix misplaced return statement in s5p_mfc_suspend()
[media] exynos-gsc: Add missing static storage class specifiers
[media] exynos-gsc: Remove <linux/version.h> header file inclusion
[media] s5p-fimc: Fix incorrect condition in fimc_lite_reqbufs()
[media] s5p-tv: Fix potential NULL pointer dereference error
[media] s5k6aa: Fix possible NULL pointer dereference
...
Rahul Sharma [Thu, 4 Oct 2012 15:18:56 +0000 (20:48 +0530)]
drm: exynos: hdmi: remove drm common hdmi platform data struct
exynos-drm-hdmi need context pointers from hdmi and mixer. These
pointers were expected from the plf data. Cleaned this dependency
by exporting i/f which are called by hdmi, mixer driver probes
for setting their context.
Rahul Sharma [Thu, 4 Oct 2012 15:18:54 +0000 (20:48 +0530)]
drm: exynos: hdmi: replace is_v13 with version check in hdmi
This patch removed the is_v13 variable from the hdmi driver context.
It is replaced with condition check for the hdmi version. This cleans
the way for handling further hdmi versions.
Rahul Sharma [Thu, 4 Oct 2012 15:18:52 +0000 (20:48 +0530)]
drm: exynos: hdmi: add support to disable video processor in mixer
This patch adds support for disabling the video processor code based
on the platform type. This is done based on a field in the mixer driver
data which changes with the platform variant.
Rahul Sharma [Thu, 4 Oct 2012 15:18:51 +0000 (20:48 +0530)]
drm: exynos: hdmi: add support for platform variants for mixer
This patch adds the support for multiple mixer versions avaialble in
various platform variants. Version is passed as a driver data field
instead of paltform data.
Rahul Sharma [Thu, 4 Oct 2012 15:18:48 +0000 (20:48 +0530)]
drm: exynos: remove drm hdmi platform data struct
This patch removes the drm hdmi platform data structure which is no
longer in use by drm hdmi driver after this patch set get merged. s5p
hdmi platform data structure is used instead.
The 'exynos-drm-hdmi' driver makes use of s5p-tv platform devices. Therefore
the driver should use the same platform data to prevent crashes caused by
dereferencing incorrect types. This patch corrects the exynos-drm-hdmi driver
to the platform data from s5p-hdmi.
Wade Farnsworth [Tue, 2 Oct 2012 16:08:30 +0000 (17:08 +0100)]
ARM: 7548/1: include linux/sched.h in syscall.h
The syscall tracing patch introduces a compile bug in lttng-modules
when the latter calls syscall_get_nr(), similar to the following:
<path-to-linux>/arch/arm/include/asm/syscall.h:21:2: error: implicit declaration of function 'task_thread_info' [-Werror=implicit-function-declaration]
The issue is that we are using task_thread_info() in the
syscall_get_nr() function in asm/syscall.h, but not explicitly
including sched.h from this file, so we can expect this bug might
surface any time that syscall_get_nr() is called.
Linus Torvalds [Fri, 5 Oct 2012 03:01:30 +0000 (12:01 +0900)]
Merge tag 'mfd-3.7-1' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6
Pull MFD changes from Samuel Ortiz:
"MFD bits for the 3.7 merge window.
As usual we have a few new drivers:
- TI LP8788
- TI OMAP USB TLL
- Maxim MAX8907
- SMSC ECE1099
- Dialog Semiconductor DA9055
- A simpler syscon driver that allow us to get rid of the anatop one.
Drivers are also gradually getting Device Tree and IRQ domain support.
The following drivers got DT support:
- palmas, 88pm860x, tc3589x and twl4030-audio
And those ones now use the IRQ domain APIs:
- 88pm860x, tc3589x, db8500_prcmu
Also some other interesting changes:
- Intel's ICH LPC now supports Lynx Point
- TI's twl4030-audio added a GPO child
- tps6527 enabled its backlight subdevice
- The twl6030 pwm driver moved to the new PWM subsystem
And finally a bunch of cleanup and casual fixes for mc13xxx, 88pm860x,
palmas, ab8500, wm8994, wm5110, max8907 and the tps65xxx family."
Fix up various annoying conflicts: the DT and IRQ domain support came in
twice and was already in 3.6. And then it was apparently rebased.
Guys, DON'T REBASE!
* tag 'mfd-3.7-1' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6: (89 commits)
ARM: dts: Enable 88pm860x pmic
mfd: 88pm860x: Move gpadc init into touch
mfd: 88pm860x: Device tree support
mfd: 88pm860x: Use irqdomain
mfd: smsc: Add support for smsc gpio io/keypad driver
backlight: tps65217_bl: Add missing platform_set_drvdata in tps65217_bl_probe
mfd: DA9055 core driver
mfd: tps65910: Add alarm interrupt of TPS65910 RTC to mfd device list
mfd: wm5110: Add register patches for revision B
mfd: wm5110: Disable control interface error report for WM5110 rev B
mfd: max8907: Remove regulator-compatible from DT docs
backlight: Add TPS65217 WLED driver
mfd: Add backlight as subdevice to the tps65217
mfd: Provide the PRCMU with its own IRQ domain
mfd: Fix max8907 sparse warning
mfd: Add lp8788 mfd driver
mfd: dbx500: Provide a more accurate smp_twd clock
mfd: rc5t583: Fix warning messages
regulator: palmas: Add DT support
mfd: palmas: Change regulator defns to better suite DT
...
Sean Hefty [Wed, 3 Oct 2012 23:42:31 +0000 (23:42 +0000)]
RDMA/cma: Check that retry count values are in range
The retry_count and rnr_retry_count connection parameters are both
3-bit values. Check that the values are in range and reduce if
they're not.
This fixes a problem reported by Doug Ledford <[email protected]>
that resulted in the userspace rping test (part of the librdmacm
samples) failing to run over Intel IB HCAs.
Signed-off-by: Sean Hefty <[email protected]>
[ Use min_t() to avoid warnings about type mismatch. - Roland ]
Al Viro [Thu, 4 Oct 2012 20:52:53 +0000 (13:52 -0700)]
sparc64: Rearrange thread info to cheaply clear syscall noerror state.
After fixing a couple of brainos, it even seems to work. What's done here
is move of ->syscall_noerror right before FPDEPTH byte in ->flags and
using sth to [%g6 + TI_SYS_NOERROR] instead of stb to [%g6 + TI_FPDEPTH] in
both branches of etrap_save. AFAICS, that ought to be solid. Again,
deciding what to do with now unused delay slot of branch on ->syscall_noerror
and dealing with the order of tests in ret_from_sys is a separate question,
but at least that way we don't have to clean ->syscall_noerror in there at
all. AFAICS, it ought to be a clear win - sth is not going to cost more than
stb on etrap_64.S side of things, and we are losing write on syscalls.S one.
projects / linux.git / log