fix __aggregate_redistribute_shares() related lockup reported by
David S. Miller.
The problem this code tries to solve is 'accurately' calculating the 'fair'
share of the group weight for each cpu. The current code falls back to a global
group rebalance in case the sched_domain's span it looks at has no shares, but
does have tasks.
The reason it gets stuck here, is because its inherently racy - if someone
steals the last task after we compute the agg->rq_weight, but before we
rebalance, we'll never get out of the loop.
We could of course go fix that, but while looking at this issue I found that
this 'fallback' wasn't nearly as rare as I'd hoped it to be. In fact its quite
common - and given it walks the whole machine, thats very bad.
The new approach is simple (why didn't I think of it before?), we set the
aggregate shares to the full task group weight, and each larger sched domain
that encounters an aggregate shares larger than the weight, clips it (it
already re-distributes anyway).
This nicely converges to the desired global picture where the sum of all
shares equals the task group weight.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
David Miller reported:
|--------------->
the following commit:
| commit 27ec440779
| Author: Ingo Molnar <mingo@elte.hu>
| Date: Thu Feb 28 21:00:21 2008 +0100
|
| sched: make cpu_clock() globally synchronous
|
| Alexey Zaytsev reported (and bisected) that the introduction of
| cpu_clock() in printk made the timestamps jump back and forth.
|
| Make cpu_clock() more reliable while still keeping it fast when it's
| called frequently.
|
| Signed-off-by: Ingo Molnar <mingo@elte.hu>
causes watchdog triggers when a cpu exits NOHZ state when it has been
there for >= the soft lockup threshold, for example here are some
messages from a 128 cpu Niagara2 box:
[ 168.106406] BUG: soft lockup - CPU#11 stuck for 128s! [dd:3239]
[ 168.989592] BUG: soft lockup - CPU#21 stuck for 86s! [swapper:0]
[ 168.999587] BUG: soft lockup - CPU#29 stuck for 91s! [make:4511]
[ 168.999615] BUG: soft lockup - CPU#2 stuck for 85s! [swapper:0]
[ 169.020514] BUG: soft lockup - CPU#37 stuck for 91s! [swapper:0]
[ 169.020514] BUG: soft lockup - CPU#45 stuck for 91s! [sh:4515]
[ 169.020515] BUG: soft lockup - CPU#69 stuck for 92s! [swapper:0]
[ 169.020515] BUG: soft lockup - CPU#77 stuck for 92s! [swapper:0]
[ 169.020515] BUG: soft lockup - CPU#61 stuck for 92s! [swapper:0]
[ 169.112554] BUG: soft lockup - CPU#85 stuck for 92s! [swapper:0]
[ 169.112554] BUG: soft lockup - CPU#101 stuck for 92s! [swapper:0]
[ 169.112554] BUG: soft lockup - CPU#109 stuck for 92s! [swapper:0]
[ 169.112554] BUG: soft lockup - CPU#117 stuck for 92s! [swapper:0]
[ 169.171483] BUG: soft lockup - CPU#40 stuck for 80s! [dd:3239]
[ 169.331483] BUG: soft lockup - CPU#13 stuck for 86s! [swapper:0]
[ 169.351500] BUG: soft lockup - CPU#43 stuck for 101s! [dd:3239]
[ 169.531482] BUG: soft lockup - CPU#9 stuck for 129s! [mkdir:4565]
[ 169.595754] BUG: soft lockup - CPU#20 stuck for 93s! [swapper:0]
[ 169.626787] BUG: soft lockup - CPU#52 stuck for 93s! [swapper:0]
[ 169.626787] BUG: soft lockup - CPU#84 stuck for 92s! [swapper:0]
[ 169.636812] BUG: soft lockup - CPU#116 stuck for 94s! [swapper:0]
It's simple enough to trigger this by doing a 10 minute sleep after a
fresh bootup then starting a parallel kernel build.
I suspect this might be reintroducing a problem we've had and fixed
before, see the thread:
http://marc.info/?l=linux-kernel&m=119546414004065&w=2
<---------------|
touch the softlockup watchdog when exiting NOHZ state - we are
obviously not locked up.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Thomas Gleixner debugged a particularly ugly seqlock related livelock:
do not process the seq-read section if we know it beforehand that the
test at the end of the section will fail ...
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Stephen Rothwell noticed that:
Commit 2be621498d ("x86: dma-ops on highmem
fix") in Linus' tree introduced a new warning (noticed in the x86_64
allmodconfig build of linux-next):
drivers/pci/intel-iommu.c:2240: warning: initialization from incompatible pointer type
Which points at an instance of map_single that needs updating.
Fix it to the new prototype.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
[CIFS] Fix typo in previous commit
[CIFS] Fix define for new proxy cap to match documentation
[CIFS] Fix UNC path prefix on QueryUnixPathInfo to have correct slash
[CIFS] Reserve new proxy cap for WAFS
[CIFS] Add various missing flags and defintions
[CIFS] make cifs_dfs_automount_list_static
[CIFS] Fix oops when slow oplock process races with unmount
[CIFS] Fix acl length when very short ACL being modified by chmod
[CIFS] Fix looping on reconnect to Samba when unexpected tree connect fail on reconnect
[CIFS] minor update to change log
Fix a regression in the RXKAD security module introduced in:
commit 91e916cffe
Author: Al Viro <viro@ftp.linux.org.uk>
Date: Sat Mar 29 03:08:38 2008 +0000
net/rxrpc trivial annotations
A variable was declared as a 16-bit type rather than a 32-bit type.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-with-apologies-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-of-by: Linus Torvalds <torvalds@linux-foundation.org>
This snuck in through 919ee677b6
("[SPARC64]: Add NUMA support")
Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.linux-nfs.org/projects/trondmy/nfs-2.6: (80 commits)
SUNRPC: Invalidate the RPCSEC_GSS session if the server dropped the request
make nfs_automount_list static
NFS: remove duplicate flags assignment from nfs_validate_mount_data
NFS - fix potential NULL pointer dereference v2
SUNRPC: Don't change the RPCSEC_GSS context on a credential that is in use
SUNRPC: Fix a race in gss_refresh_upcall()
SUNRPC: Don't disconnect more than once if retransmitting NFSv4 requests
SUNRPC: Remove the unused export of xprt_force_disconnect
SUNRPC: remove XS_SENDMSG_RETRY
SUNRPC: Protect creds against early garbage collection
NFSv4: Attempt to use machine credentials in SETCLIENTID calls
NFSv4: Reintroduce machine creds
NFSv4: Don't use cred->cr_ops->cr_name in nfs4_proc_setclientid()
nfs: fix printout of multiword bitfields
nfs: return negative error value from nfs{,4}_stat_to_errno
NLM/lockd: Ensure client locking calls use correct credentials
NFS: Remove the buggy lock-if-signalled case from do_setlk()
NLM/lockd: Fix a race when cancelling a blocking lock
NLM/lockd: Ensure that nlmclnt_cancel() returns results of the CANCEL call
NLM: Remove the signal masking in nlmclnt_proc/nlmclnt_cancel
...
* 'for-linus' of git://linux-nfs.org/~bfields/linux: (52 commits)
knfsd: clear both setuid and setgid whenever a chown is done
knfsd: get rid of imode variable in nfsd_setattr
SUNRPC: Use unsigned loop and array index in svc_init_buffer()
SUNRPC: Use unsigned index when looping over arrays
SUNRPC: Update RPC server's TCP record marker decoder
SUNRPC: RPC server still uses 2.4 method for disabling TCP Nagle
NLM: don't let lockd exit on unexpected svc_recv errors (try #2)
NFS: don't let nfs_callback_svc exit on unexpected svc_recv errors (try #2)
Use a zero sized array for raw field in struct fid
nfsd: use static memory for callback program and stats
SUNRPC: remove svc_create_thread()
nfsd: fix comment
lockd: Fix stale nlmsvc_unlink_block comment
NFSD: Strip __KERNEL__ testing from unexported header files.
sunrpc: make token header values less confusing
gss_krb5: consistently use unsigned for seqnum
NFSD: Remove NFSv4 dependency on NFSv3
SUNRPC: Remove PROC_FS dependency
NFSD: Use "depends on" for PROC_FS dependency
nfsd: move most of fh_verify to separate function
...
* git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/v4l-dvb: (452 commits)
V4L/DVB (7731): tuner-xc2028: fix signal strength calculus
V4L/DVB (7730): tuner-xc2028: Fix SCODE load for MTS firmwares
V4L/DVB (7729): Fix VIDIOCGAP corruption in ivtv
V4L/DVB (7728): tea5761: bugzilla #10462: tea5761 autodetection code were broken
V4L/DVB (7726): cx23885: Enable cx23417 support on the HVR1800
V4L/DVB (7725): cx23885: Add generic cx23417 hardware encoder support
V4L/DVB (7723): pvrusb2: Clean up input selection list generation in V4L interface
V4L/DVB (7722): pvrusb2: Implement FM radio support for Gotview USB2.0 DVD 2
V4L/DVB (7721): pvrusb2: Restructure cx23416 firmware loading to have a common exit point
V4L/DVB (7720): pvrusb2: Fix bad error code on cx23416 firmware load failure
V4L/DVB (7719): pvrusb2: Implement input selection enforcement
V4L/DVB (7718): pvrusb2-dvb: update Kbuild selections
V4L/DVB (7717): pvrusb2-dvb: add DVB-T support for Hauppauge pvrusb2 model 73xxx
V4L/DVB (7716): pvrusb2: clean up global functions
V4L/DVB (7715): pvrusb2: Clean out all use of __FUNCTION__
V4L/DVB (7714): pvrusb2: Fix hang on module removal
V4L/DVB (7713): pvrusb2: Implement cleaner DVB kernel thread shutdown
V4L/DVB (7712): pvrusb2: Close connect/disconnect race
V4L/DVB (7711): pvrusb2: Fix race on module unload
V4L/DVB (7710): pvrusb2: Implement critical digital streaming quirk for onair devices
...
RFC 2203 requires the server to drop the request if it believes the
RPCSEC_GSS context is out of sequence. The problem is that we have no way
on the client to know why the server dropped the request. In order to avoid
spinning forever trying to resend the request, the safe approach is
therefore to always invalidate the RPCSEC_GSS context on every major
timeout.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
There are different tables for MTS firmwares. This should be taken into
account while selecting the proper firmware.
While at tuner-xc2028.h, improve some comments.
Thanks to Edward J. Sheldrake <ejs1920@yahoo.co.uk> for helping to
diagnose such troubles with PAL/I standard.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Frank Bennett reported that ivtv was causing skype to crash. With help
from one of their developers he showed it was a kernel problem.
VIDIOCGCAP copies a name into a fixed length buffer - ivtv uses names
that are too long and does not truncate them so corrupts a few bytes of
the app data area.
Possibly the names also want trimming but for now this should fix the
corruption case.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
cx23885: Enable cx23417 support on the HVR1800
Signed-off-by: Steven Toth <stoth@hauppauge.com>
Signed-off-by: Michael Krufky <mkrufky@linuxtv.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Change how list of possible pvrusb2 inputs is generated to include
only those interfaces that make sense for the interface instance.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
These changes are required with the addition of digital television support
for the Hauppauge HVR1900 & HVR1950, the OnAir Creator and Sasem USB HDTV
Signed-off-by: Michael Krufky <mkrufky@linuxtv.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
This patch contains the following cleanups:
- make the following needlessly global function static:
- pvr2_hdw_set_cur_freq()
- #if 0 the following unused global functions:
- pvr2_hdw_get_state_name()
- pvr2_hdw_get_debug_info_unlocked()
- pvr2_hdw_get_debug_info_locked()
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Apparently the kernel developers no longer consider it proper
etiquette to use __FUNCTION__; everyone must instead use __func__
(even though it breaks with older compilers). And worse still, actual
effort is being expended to sweep this change throughout the kernel
source tree. Don't these people have better things to do? So...
Completely clean out all use of __FUNCTION__ from the pvrusb2 driver
(it was just in the sysfs interface). I'm not going to use __func__
either. So there.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
The pvrusb2 driver was getting had by this scenario:
1. Task A calls kthread_stop() for task B.
2. Before exiting, then Task B calls kthread_stop() for task C.
The problem is, kthread_stop() wants to allocate an internal resource
to itself (i.e. acquire a lock), which won't be released until
kthread_stop() returns. But kthread_stop() won't return until task B
is dead. But task B won't die until it finishes its call to
kthread_stop() for task C, and that will block waiting on the resource
already allocated inside task A. Deadlock.
With the pvrusb2 driver, task A is the caller to pvr_exit(), task B is
the control thread run inside of pvrusb2-context.c, and task C is any
worker thread run inside of pvrusb2-hdw.c.
This problem got introduced by the previous threading setup change,
which was itself an attempt to fix a module tear-down race (which it
actually did fix). The lesson here is that a task being waited on as
part of a kthread_stop() simply cannot be allow to also issue a
kthread_stop() - or we make sure not to issue the enclosing
kthread_stop() until we know that the inner kthread_stop() has
completed first. The solution for the pvrusb2 driver is some hackish
code which changes the main control thread tear down into a two step
process. This then makes it possible to delay issuing the
kthread_stop() on the control thread until after we know that
everything has been torn down first. (And yes, we really need that
kthread_stop() because it's the only way to safely guarantee that a
module-referencing kernel thread has safely returned back out of the
module before we finally remove the module.)
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Earlier fix to handle DVB feed thread aborts was overly-aggressive.
We can take better advantage of what kthread_stop() can do. This
change simplifies things.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
If a disconnect happens before initialization is completed, the
pvrusb2 driver can accidentally touch dangling pointers. The whole
initialization function must be protected by the big_lock, and once
inside that lock, the initialization function should abort if it is
discovered that a disconnect has already taken place.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
The pvrusb2 driver - for basically forever - was not enforcing a
proper module tear-down. Kernel threads are used inside the driver
and all must be gone before the module can be safely removed. This
changeset reimplements a chunk of pvrusb2-context.c to enforce this
correctly. Unfortunately this is not a simple fix. The new
implementation also cuts back on kernel thread usage; instead of there
being 1 control thread per instance now it's just 1 control thread
shared by all instances. (By dropping to a single thread then the
module exit function can block on its shutdown and the thread itself
can monitor and cleanly shut down all of the other instances first.)
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Implement timed measurement of encoder operation for the first time it
is run. This allows the driver to note when the encoder has been run
successfully for at least 1/4 second. On top of that implement
various bits to ensure that the encoder has been run once before
digital streaming for OnAir devices. This is done via several core
state machine tweaks.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Some tuners seem to not work in digital mode unless the encoder is
healthy. Implement a device attribute to represent this flag and
modify the core state machines to enforce this requirement.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
If the device fails to stream, the feed thread will block forever
waiting for buffers. But while in this state it was not looking for
an exit condition from the driver DVB interface. This caused the
thread to jam. Implement a new stop flag (which will be set
appropriately) to tell the thread to stop.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
When the DVB interface is not compiled, the pvr2_dvb_props struct is
not available - so it really should be ifdef'ed out as well. This
didn't cause an error because in this context its usage was as an
opaque pointer. But it really shouldn't be present at all if DVB is
not enabled.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
The commands to start / stop USB streaming for an analog device are
fairly standard, owing to the fact that all supported devices
apparently started from the same common reference design. However
with digital mode, the commands seem to vary by vendor. This change
makes that variance more explicit. It also cleans up a related
problem for OnAir devices which prevented digital mode from working at
all.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Numerous places in the driver need to issue simple commands to the FX2
microcontroller (e.g. only 1 or 2 bytes, no reply needed). Previously
each place that did this, had to take lock, set up a central buffer,
and call the function to perform the handshake. This change puts
these steps into a single spot. This also has the effect of removing
the need to mess with the control lock from numerous places in the
code.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Implement a mechanism in the pvrusb2 driver for gathering statistics
on the stream buffering, including bytes transferred, buffers handled,
buffers in flight, etc. This is useful for debugging certain classes
of streaming issues and for determining if the buffer pool size is
generally correct for the driver.
Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Peter Zijlstra