- 15 Jan, 2005 40 commits
-
-
Roland McGrath authored
When kill is used to force a core dump, __group_complete_signal uses the group_stop_count machinery to stop other threads from doing anything more before the signal-taking thread starts the coredump synchronization. This intentionally results in group_stop_count always still being > 0 when the signal-taking thread gets into do_coredump. However, that has the unintended effect that signal_pending can return true when called from the filesystem code while writing the core dump file. For NFS mounts using the "intr" option, this results in NFS operations bailing out before they even try, so core files never get successfully dumped on such a filesystem when the crash was induced by an asynchronous process-wide signal. This patch fixes the problem by clearing group_stop_count after the coredump synchronization is complete. The locking I threw in is not directly related, but always should have been there and may avoid some potential races with kill. Signed-off-by:
Roland McGrath <roland@redhat.com> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Prasanna Meda authored
Since Roland changed now to wakeup tracee with kill, I guess this needs to be fixed. http://linus.bkbits.net:8080/linux-2.5/gnupatch@41e3fe5fIRH-W3aDnXZgfQ-qIvuXYg ptrace_setsiginfo/ptrace_getsiginfo need to do locked access to last_siginfo. ptrace_notify()/ptrace_stop() sets the current->last_siginfo and sleeps on schedule(). It can be waked up by kill signal from signal_wake_up before debugger wakes it up. On return from schedule(), the current->last_siginfo is reset. Signed-off-by:
Prasanna Meda <pmeda@akamai.com> Acked-by:
-
Roland McGrath authored
If one thread uses ptrace on another thread in the same thread group, there can be a deadlock when calling exec. The ptrace_stop change ensures that no tracing stop can be entered for a queued signal, or exit tracing, if the tracer is part of the same dying group. The exit_notify change prevents a ptrace zombie from sticking around if its tracer is in the midst of a group exit (which an exec fakes), so these zombies don't hold up de_thread's synchronization. The de_thread change ensures the new thread group leader doesn't wind up ptracing itself, which would produce its own deadlocks. Signed-off-by:
-
Roland McGrath authored
When threads are sharing mm via CLONE_VM (linuxthreads, vfork), there is a race condition where one thread doing a core dump and synchronizing all mm-sharing threads for it can deadlock waiting for another thread that just did an exec and will never synchronize. This patch makes the exec_mmap check for a pending core dump and punt the exec to synchronize with that, as if the core dump had struck before entering the execve system call at all. Signed-off-by:
-
Roland McGrath authored
In the oddball situation where one thread is using ptrace on another thread sharing the same mm, and then someone sharing that mm causes a coredump, there is a deadlock possible if the traced thread is in TASK_TRACED state. It leaves all the threads sharing that mm wedged and permanently unkillable. This patch checks for that situation and brings a thread out of TASK_TRACED if its tracer is part of the same coredump (i.e. shares the same mm). It's not pretty, but it does the job. Signed-off-by:
-
Michael S. Tsirkin authored
To make life bearable for out-of kernel modules, the following patch adds 2 macros so that existance of unlocked_ioctl and ioctl_compat can be easily detected. Signed-off-by:
Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by:
-
Christoph Hellwig authored
- add ->unlocked_ioctl method and a do_ioctl wrapper in ioctl.c so all places calling ->ioctl get it. THis provides us a patch to migrate away from holding bkl across ioctl implementations. - add ->compat_ioctl method and call it in compat_sys_ioctl before doing the hash lookup for registered handlers. - streamline compat_sys_ioctl and move the complex error reporting into a function of its own From: "Michael S. Tsirkin" <mst@mellanox.co.il> Handle generic ioctl commands by falling back on static conversion functions in fs/compat_ioctl.c on -ENOIOCTLCMD code. From: "Michael S. Tsirkin" <mst@mellanox.co.il> With new unlocked_ioctl and ioctl_compat, ioctls can now be as fast as read/write. So lets use fget_light/fput_light there, to get some speedup in common case on SMP. Signed-off-by:
-
Andreas Gruenbacher authored
This started of as a patch by Alex Tomas <alex@clusterfs.com> and got an overhaul by me. The on-disk structure used is the same as in Alex's original patch. Signed-off-by:
Andreas Gruenbacher <agruen@suse.de> Signed-off-by:
-
Andreas Gruenbacher authored
The in_mem optimization in ext3_get_inode_loc avoids a disk read when only the requested inode in the block group is allocated: In that case ext3_get_inode_loc assumes that it can recreate the inode from the in-memory inode. This is incorrect with in-inode extended attributes, which don't have a shadow copy in memory. Hide the in_mem option and clarify comments; the subsequent ea-in-inode changes the in_mem check as required. Signed-off-by:
-
Andreas Gruenbacher authored
Clean up several things in the xattr code, and prepare it for in-inode attributes: * Add the ext3_xattr_check_names, ext3_xattr_check_block, and ext3_xattr_check_entry functions for checking xattr data structures. * Add the ext3_xattr_find_entry, ext3_xattr_list_entries, and ext3_xattr_set_entry functions for manipulating xattr entries. Switch to using these functions in ext3_xattr_get, ext3_xattr_list, and ext3_xattr_set_handle. * Merge ext3_xattr_set_handle and ext3_xattr_set_handle2. * Rename the HDR and FIRST_ENTRY macros. * We have no way to deal with a ext3_xattr_cache_insert failure, so make it return void. * Make the debug messages more useful. Signed-off-by:
-
Andreas Gruenbacher authored
The ext3_xattr_set_handle2 and ext3_xattr_delete_inode functions contain duplicate code to decrease the reference count of an xattr block. Move this to a separate function. Also we know we have exclusive access to the inode in ext3_xattr_delete_inode; there is no need to grab the xattr_sem lock. Signed-off-by:
-
Andreas Gruenbacher authored
The ext3_xattr_set_handle2 and ext3_xattr_delete_inode functions contain duplicate code to decrease the reference count of an xattr block. Move this to a separate function. Also we know we have exclusive access to the inode in ext3_xattr_delete_inode; there is no need to grab the xattr_sem lock. Signed-off-by:
-
Andreas Gruenbacher authored
The use of journal_release_buffer is unsafe; it can overflow the journal: When a buffer is stolen from a transaction and later removed from that transaction with journal_release_buffer, the buffer is not accounted to the transaction that now "owns" the buffer, and one extra credit appears to be available. Don't use journal_release_buffer: We did rely on the buffer lock to synchronize xattr block accesses, and get write access to the buffer first to get atomicity. Return the mb_cache_entry from ext3_xattr_cache_find instead, and do the check/update under its lock. Only get write access when we know we will use the buffer. Signed-off-by:
-
Andreas Gruenbacher authored
Andrew Tridgell and Stephen C. Tweedie have reported two different Oopses caused by a race condition in the mbcache, which is responsible for extended attribute sharing in ext2 and ext3. Stephen tracked down the bug; I did the fix. Explanation: The mbcache caches the locations and content hashes of xattr blocks. There are two access strategies: [1] xattr block disposal via mb_cache_entry_get(), [2] xattr block reuse (sharing) via mb_cache_entry_find_{first,next}(). There is no locking between the two methods, so between one mb_cache_entry_find_x and the next, a mb_cache_entry_get might come in, unhash the cache entry, and change the journaling state of the xattr buffer. Subsequently, two things can happen: [a] the next mb_cache_entry_find_x may try to follow the mbcache hash chain starting from the entry that has become unhashed, which now is a stale pointer, [b] the block may have become deallocated, and then we try to reuse it. Fix this by converting the mbcache into a readers-writer style lock, and protect all block accesses in ext2/ext3 by the mbcache entry lock. This ensures that destroying blocks is an exclusive operation that may not overlap xattr block reuse, while allowing multiple "re-users". Write access to the xattr block's buffer is protected by the buffer lock. Signed-off-by: -
Andreas Gruenbacher authored
There is no need to export struct mb_cache outside mbcache.c. Move struct mb_cache to fs/mbcache.c and remove the superfluous struct mb_cache_entry_index declaration. Signed-off-by:
-
Andreas Gruenbacher authored
Revert the recently-added (post-2.6.10) ea-in-inode speedup patch. We have a new one. Signed-off-by:
-
Markus Lidel authored
- removed commented part in header which is not used anymore - added wrappers for printk to make code more readable - cleaned up logging output and removed unneccessary output Signed-off-by:
Markus Lidel <Markus.Lidel@shadowconnect.com> Signed-off-by:
-
Markus Lidel authored
- I2O Block OSM contained a possible race condition (by Wolfram Joost) - In pci.c the memory area which will be synced was wrong (original from Wolfram Joost) - number of OSM's which could be loaded is increased from 4 to 8 Signed-off-by:
-
Randy Dunlap authored
swiotlb: Fix gcc printk format warning on x86_64, OK for ia64: arch/ia64/lib/swiotlb.c:351: warning: long unsigned int format, long long unsigned int arg (arg 2) Signed-off-by:
Randy Dunlap <rddunlap@osdl.org> Signed-off-by:
-
Jörn Engel authored
Fixes to block2mtd.c - the previous driver was only compile-tested: o Corrected copyright statements and some comments. o Renamed blockmtd to block2mtd. o Fixed the newline handling, it oopsed the kernel if no erase size was passed. o Replaced some printk() by INFO(). o Tried (but failed) to replace the readahead code. Signed-off-by:
Jörn Engel <joern@wohnheim.fh-wedel.de> Signed-off-by:
-
Alan Cox authored
Fixes for various smbfs data leak bugs from Alan, Chuck Ebbert and various people on various mailing lists. Signed-off-by:
-
Pawel Sikora authored
This patch kills tons of gcc4 warnings: pointer targets in passing argument 2 of 'csum_and_copy_from_user' differ in signedness From: Hirokazu Takata <takata@linux-m32r.org> * include/asm-m32r/checksum.h: build fix * arch/m32r/lib/csum_partial_copy.c: - update comment - cosmetic changes: change indentation Signed-off-by:
Hirokazu Takata <takata@linux-m32r.org> Signed-off-by:
-
Stefan Bader authored
Disable seek on z/VM log reader misc device by using nonseekable_open(). Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by:
-
Carsten Otte authored
Make the ECKD compatible disk layout labling detection conditional to run only on ECKD compatible disk layout volumes, do a fall back into the default LNX/unlabled case otherwise. Signed-off-by:
-
Horst Hummel authored
dasd driver changes: - Fix debug feature usage. The sprinf event/exception functions write a pointer to a format string (%s) to the log. These strings must be valid at the time the DBF-log is read. - Some coding style reformatting. Signed-off-by:
-
Ursula Braun-Krahl authored
Remove the irq_exit call on error path in iucv_irq_handler. irq_exit is done in do_extint(). Signed-off-by:
-
Martin Schwidefsky authored
3270 console changes: - Initialize timer element before first use. Signed-off-by:
-
Steffen Thoss authored
Common I/O layer changes: - Check if AIF is available on hardware before enabling the AIF time delay disablement facility. Signed-off-by:
-
Heiko Carstens authored
Collaborative memory management inferface changes: - Allow cmmthread to run on every cpu. Signed-off-by:
-
Heiko Carstens authored
s390 core changes: - Fix mm_struct leak on cpu hotplug. - Improved cpu detection logic to avoid long delay at system start. - Call cpu_relax() in cpu hotplug wait loop. - Remove #define of account_system_vtime for CONFIG_VIRT_CPU_ACCOUNTING=n. - Regenerate default configuration. Signed-off-by:
-
Christian Bornträger authored
S/390 needs this for its binfmt_elf32 module. Signed-off-by:
Christian Borntraeger <cborntra@de.ibm.com> Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
It seems that linker script for userspace software are quite toolchain-depending, at least because what we use is a merge between builtin LD scripts (see strings /usr/bin/ld) and normal kernel linking scripts. Plus, a number of people are having toolchain-related troubles building UML (even assertion failures on linking, with Gentoo and Fedora 2). So, let's try to make UML nicer for binutils. Signed-off-by:
Paolo 'Blaisorblade' Giarrusso <blaisorblade_spam@yahoo.it> Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
From: Bodo Stroesser <bstroesser@fujitsu-siemens.com> Add stack addresses to print of symbols from stack trace. For stack analysis it's important to have this information. Signed-off-by:
Bodo Stroesser <bstroesser@fujitsu-siemens.com> For UML, we should also copy the CONFIG_FRAME_POINTER stack walking from i386, and move the result to sys-i386. Another note: this should be done for i386 also, if ksymoops does not have problems. Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
Copy some code from i386 to print the stack content. Rough form yet, should work although. Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
* Fix the use of errno: it refers to the __errno_location glibc definition when in ubd_user.c, and hence works; but in ubd_kern.c it refers to kernel_errno, which is different. So use the return value of os_* functions, as we should always have done. * Remove {read,write}_ubd_fs(), which are just silly. Signed-off-by: -
Paolo \'Blaisorblade\' Giarrusso authored
Most code of ubd_user.c already uses the os_* functions, so it can be moved to ubd_kern.c. This patch simply moves the code without any hidden changes. The only change is inside io_thread(): since it calls signal(), I created a little function in ubd_user.c which just calls signal() with the right parameters. In a later patch (send together) I'll do some changes, to fix the usage of errno (which makes this code break when moved in a kernelspace file) and for some other little cleanups. Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
When parsing the <flags> section in ubd<n><flags>=file[,file2], instead of requiring that the flags are specified in a certain order, just make the code smarter. Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
Currently we can use this syntax ubd<n>[<flags>]=file1,file2. However, writing things as ubd0=~/Uml/file1,~/Uml/file2 does not work; in fact, the shell won't expand the second '~', since it's not at a path beginning; possibly even other shell expansions don't work here. So simply allow using, instead of the ',' separator, the ':' separator. The ',' separator can still be used to preserve backward compatibility. Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
Return an early error message when no TT support is compiled in and no SKAS support is detected. Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
UML has his own initcall mechanism to handle his special userspace initialization (they are called in different moments, so they are indeed It must also duplicate some definition for the benefit of userspace code - but those definition weren't in sync with the main code. Also, the UML own macros missed __attribute_used__. Both problems are fixed by this patch. Signed-off-by:
-
