- 27 Jul, 2005 40 commits
-
-
Peter Staubach authored
I believe that there is a problem with the handling of POSIX locks, which the attached patch should address. The problem appears to be a race between fcntl(2) and close(2). A multithreaded application could close a file descriptor at the same time as it is trying to acquire a lock using the same file descriptor. I would suggest that that multithreaded application is not providing the proper synchronization for itself, but the OS should still behave correctly. SUS3 (Single UNIX Specification Version 3, read: POSIX) indicates that when a file descriptor is closed, that all POSIX locks on the file, owned by the process which closed the file descriptor, should be released. The trick here is when those locks are released. The current code releases all locks which exist when close is processing, but any locks in progress are handled when the last reference to the open file is released. There are three cases to consider. One is the simple case, a multithreaded (mt) process has a file open and races to close it and acquire a lock on it. In this case, the close will release one reference to the open file and when the fcntl is done, it will release the other reference. For this situation, no locks should exist on the file when both the close and fcntl operations are done. The current system will handle this case because the last reference to the open file is being released. The second case is when the mt process has dup(2)'d the file descriptor. The close will release one reference to the file and the fcntl, when done, will release another, but there will still be at least one more reference to the open file. One could argue that the existence of a lock on the file after the close has completed is okay, because it was acquired after the close operation and there is still a way for the application to release the lock on the file, using an existing file descriptor. The third case is when the mt process has forked, after opening the file and either before or after becoming an mt process. In this case, each process would hold a reference to the open file. For each process, this degenerates to first case above. However, the lock continues to exist until both processes have released their references to the open file. This lock could block other lock requests. The changes to release the lock when the last reference to the open file aren't quite right because they would allow the lock to exist as long as there was a reference to the open file. This is too long. The new proposed solution is to add support in the fcntl code path to detect a race with close and then to release the lock which was just acquired when such as race is detected. This causes locks to be released in a timely fashion and for the system to conform to the POSIX semantic specification. This was tested by instrumenting a kernel to detect the handling locks and then running a program which generates case #3 above. A dangling lock could be reliably generated. When the changes to detect the close/fcntl race were added, a dangling lock could no longer be generated. Cc: Matthew Wilcox <willy@debian.org> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Christian Borntraeger authored
vmcp_write uses GPF_DMA for the memory allocation of the response buffer, so it can use the low level function __cpcmd directly, no need to call the wrapper. Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by:
-
Heiko Carstens authored
Fix wrong move direction of timer values for cpu accounting in case of a machine check that indicates a broken cpu timer. Signed-off-by:
-
Heiko Carstens authored
Limit reported memory size to 2GB if running in 31 bit mode. Signed-off-by:
-
Stefan Bader authored
Tape driver fixes: - Added deferred condition handling to tape driver core. - Added ability to handle busy conditions. - Code cleanup. Signed-off-by:
-
Horst Hummel authored
Free dasd slab cache on module unload. Signed-off-by:
-
Horst Hummel authored
The FBA discipline does not use retries for failed requests. A request fails after the first unsuccessful start attempt. There are some rare conditions (e.g. CIO path recovery) in which the start of an i/o on a fba device can fail. A tiny amount of retries is therefore reasonable. Signed-off-by:
-
Cornelia Huck authored
When processing resource accessibility events, continue searching for further affected subchannels if a link address is provided in the event information. Signed-off-by:
-
Cornelia Huck authored
Fix debug data in case of an interface-control or channel-control check: don't log the not yet accumulated interrupt-response-block, but the one we just received. Signed-off-by:
-
Heiko Carstens authored
The kernel uses the SIGP external call order code to signal other CPUs. When running with dedicated CPUs external calls don't get delivered immediately but within a fixed polling invervall. This can lead to delays where the system appears to do nothing. Replace the SIGP external call order with the SIGP emergency call order since this one gets delivered immediately. Signed-off-by:
-
Heiko Carstens authored
The atomic64 primitives are supposed to have 64-bit parameters instead of int. Signed-off-by:
-
Martin Schwidefsky authored
The find_next_{zero}_bit primitives on s390* should never return a bit number bigger then the bit field size. In the case of a bitfield that doesn't end on a word boundary, an offset that makes the search start at the last word of the bit field and the last word doesn't contain any zero/one bits the search is continued with a call to find_first_bit with a negative size. The search normally ends pretty quickly because the words following the bit field contain a mix of zeros and ones. But the bit number that is returned in this case is too big. To fix this and additional if to check for this case is needed. To make the code easier to read I removed the assembler parts from the find_next_{zero}_bit functions, the C-ified code is as good. Signed-off-by: -
Martin Schwidefsky authored
Split spin lock and r/w lock implementation into a single try which is done inline and an out of line function that repeatedly tries to get the lock before doing the cpu_relax(). Add a system control to set the number of retries before a cpu is yielded. The reason for the spin lock retry is that the diagnose 0x44 that is used to give up the virtual cpu is quite expensive. For spin locks that are held only for a short period of time the costs of the diagnoses outweights the savings for spin locks that are held for a longer timer. The default retry count is 1000. Signed-off-by:
-
Miles Bader authored
Signed-off-by:
Miles Bader <miles@gnu.org> Signed-off-by:
-
Miles Bader authored
These changes are untested (I no longer have the hardware). Signed-off-by:
-
Miles Bader authored
Signed-off-by:
-
Miles Bader authored
Signed-off-by:
-
Miles Bader authored
Signed-off-by:
-
Miles Bader authored
Signed-off-by:
-
Miles Bader authored
Signed-off-by:
-
Mikael Starvik authored
* Added abstraction layer for subarchs. * Added v32 support. * Renamed driver. Signed-off-by:
Mikael Starvik <starvik@axis.com> Acked-by:
Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> Signed-off-by:
-
Mikael Starvik authored
New CRIS sub architecture named v32. From: Dave Jones <davej@redhat.com> Fix swapped kmalloc args Signed-off-by:
Dave Jones <davej@redhat.com> Signed-off-by:
-
Mikael Starvik authored
Patches to make CRIS work with 2.6.12. Signed-off-by:
-
Mikael Starvik authored
Include file for synchronous serial port driver. Signed-off-by:
-
Mikael Starvik authored
Patches to support SMP. * Each CPU has its own current_pgd. * flush_tlb_range is implemented as flush_tlb_mm. * Atomic operations implemented with spinlocks. * Semaphores implemented with spinlocks. Signed-off-by:
-
Mikael Starvik authored
System-level profiler. Signed-off-by:
-
Mikael Starvik authored
Patches to make it possible to add PCI support. Signed-off-by:
-
Mikael Starvik authored
Memory management patches. * SMP support. * Non-executable stack (on v32). * 4-level page tables. * Added simple Thread Local Storage support. Signed-off-by:
-
Mikael Starvik authored
* Start threads with IRQs enabled. * Move symbol exports to arch specific file. * Prepare for real command line in the future. * Handle csum for partition that crosses flash boundary. * Set utsname. Signed-off-by:
-
Mikael Starvik authored
Use the generic IRQ framework Signed-off-by:
-
Mikael Starvik authored
Added I/O and DMA allocators to be used by drivers. Signed-off-by:
-
Mikael Starvik authored
Updates to device drivers. * Use I/O and DMA allocators. * Use wait_event_interruptible instead of interrutiple_sleep_on. * Added spinlocks SMP. * Changed restore_flags to local_irq_restore etc. * Updated IDE driver include to fit 2.6.12. Signed-off-by:
-
Mikael Starvik authored
Improvements to crash debug code. Signed-off-by:
-
Mikael Starvik authored
Changes to console. * Added LF->CRLF translation * Make use of real console framework. Signed-off-by:
-
Mikael Starvik authored
Changes to configuration and build system. * Added v32 sub architecture. * Use generic hard IRQ. * Added SMP options. * Added options to OOPS at NMI and reboot at OOM. * Made it possible to set objtree. * Added option to select Kernel GDB serial port. * Corrected Kconfig usage. * Added system profiler. Signed-off-by:
-
Mikael Starvik authored
Changes necessary to make the sub-arch split complete. Signed-off-by:
-
Adrian Bunk authored
There's no help text for CONFIG_DEBUG_STACKOVERFLOW - add one. Signed-off-by:
Jesper Juhl <juhl-lkml@dif.dk> Signed-off-by:
Adrian Bunk <bunk@stusta.de> Cc: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by:
-
Adrian Bunk authored
There's no help text for CONFIG_DEBUG_STACKOVERFLOW - add one. Signed-off-by:
-
Yoichi Yuasa authored
This patch has removed obsolete GIU driver for vr41xx. This patch already has been applied to mips tree. Signed-off-by:
Yoichi Yuasa <yuasa@hh.iij4u.or.jp> Cc: Ralf Baechle <ralf@linux-mips.org> Signed-off-by:
-
David Gibson authored
Presently the LparMap, one of the structures the kernel shares with the legacy iSeries hypervisor has a fixed offset address in head.S. This patch changes this so the LparMap is a normally initialized structure, without fixed address. This allows us to use macros to compute some of the values in the structure, which wasn't previously possible because the assembler always uses signed-% which gets the wrong answers for the computations in question. Unfortunately, a gcc bug means that doing this requires another structure (hvReleaseData) to be initialized in asm instead of C, but on the whole the result is cleaner than before. Signed-off-by:
David Gibson <dwg@au1.ibm.com> Signed-off-by:
-
