- 01 May, 2005 40 commits
-
-
Paul E. McKenney authored
Add a deprecated_for_modules macro that allows symbols to be deprecated only when used by modules, as suggested by Andrew Morton some months back. Signed-off-by:
Paul E. McKenney <paulmck@us.ibm.com> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Paul E. McKenney authored
The gpl exports need to be put back. Moving them to GPL -- but in a measured manner, as I proposed on this list some months ago -- is fine. Changing these particular exports precipitously is most definitely -not- fine. Here is my earlier proposal: http://marc.theaimsgroup.com/?l=linux-kernel&m=110520930301813&w=2 See below for a patch that puts the exports back, along with an updated version of my earlier patch that starts the process of moving them to GPL. I will also be following this message with RFC patches that introduce two (EXPORT_SYMBOL_GPL) interfaces to replace synchronize_kernel(), which then becomes deprecated. Signed-off-by: <paulmck@us.ibm.com> Signed-off-by:
-
Daniel Drake authored
The current logic assumes that a /proc/<PID>/task directory should have a hardlink count of 3, probably counting ".", "..", and a directory for a single child task. It's fairly obvious that this doesn't work out correctly when a PID has more than one child task, which is quite often the case. Signed-off-by:
Daniel Drake <dsd@gentoo.org> Signed-off-by:
-
Daniel Drake authored
The pid directories in /proc/ currently return the wrong hardlink count - 3, when there are actually 4 : ".", "..", "fd", and "task". This is easy to notice using find(1): cd /proc/<pid> find In the output, you'll see a message similar to: find: WARNING: Hard link count is wrong for .: this may be a bug in your filesystem driver. Automatically turning on find's -noleaf option. Earlier results may have failed to include directories that should have been searched. http://bugs.gentoo.org/show_bug.cgi?id=86031 I also noticed that CONFIG_SECURITY can add a 5th: attr, and performed a similar fix on the task directories too. Signed-off-by:
-
Stas Sergeev authored
The attached patch moves the IRQ-related SA_xxx flags (namely, SA_PROBE, SA_SAMPLE_RANDOM and SA_SHIRQ) from all the arch-specific headers to linux/signal.h. This looks like a left-over after the irq-handling code was consolidated. The code was moved to kernel/irq/*, but the flags are still left per-arch. Right now, adding a new IRQ flag to the arch-specific header, like this patch does: http://cvs.sourceforge.net/viewcvs.py/*checkout*/alsa/alsa-driver/utils/patches/pcsp-kernel-2.6.10-03.diff?rev=1.1 no longer works, it breaks the compilation for all other arches, unless you add that flag to all the other arch-specific headers too. So I think such a clean-up makes sense. Signed-off-by:
Stas Sergeev <stsp@aknet.ru> Signed-off-by:
-
Matt Mackall authored
Arrange for all kernel printks to be no-ops. Only available if CONFIG_EMBEDDED. This patch saves about 375k on my laptop config and nearly 100k on minimal configs. Signed-off-by:
Matt Mackall <mpm@selenic.com> Signed-off-by:
-
Matt Mackall authored
Remove PAGE_BUG - repalce it with BUG and BUG_ON. Signed-off-by:
-
Matt Mackall authored
This patch eliminates all kernel BUGs, trims about 35k off the typical kernel, and makes the system slightly faster. Signed-off-by:
-
Matt Mackall authored
Add a pair of rlimits for allowing non-root tasks to raise nice and rt priorities. Defaults to traditional behavior. Originally written by Chris Wright. The patch implements a simple rlimit ceiling for the RT (and nice) priorities a task can set. The rlimit defaults to 0, meaning no change in behavior by default. A value of 50 means RT priority levels 1-50 are allowed. A value of 100 means all 99 privilege levels from 1 to 99 are allowed. CAP_SYS_NICE is blanket permission. (akpm: see http://www.uwsg.iu.edu/hypermail/linux/kernel/0503.1/1921.html for tips on integrating this with PAM). Signed-off-by:
Ingo Molnar <mingo@elte.hu> Signed-off-by:
-
Cornelia Huck authored
Synchronize documentation with current interface. Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by:
-
Cornelia Huck authored
The ioctl32_conversion routines will be deprecated: Remove them from the crypto driver. Signed-off-by:
-
Cornelia Huck authored
The ioctl32_conversion routines will be deprecated: Remove them from dasd_cmb and handle the three cmb ioctls like all other dasd ioctls. Signed-off-by:
-
Horst Hummel authored
The first blocks on a cdl formatted dasd device are smaller than the blocksize of the device. Read requests are padded with a 'e5' pattern. Write requests should not pad the (user) buffer with 'e5' because a write request is not allowed to modify the buffer. Signed-off-by:
-
Stefan Weinhuber authored
The DASD device driver never reorders the I/O requests and relies on the hardware to write all data to nonvolatile storage before signaling a successful write. Hence, the only thing we have to do to support write barriers is to set the queue ordered flag. Signed-off-by:
-
Horst Hummel authored
The independent read-only flags in devmap, dasd_device and gendisk are not kept in sync. Use one bit per feature in the dasd driver and keep that bit in sync with the gendisk bit. Signed-off-by:
-
Michael Holzheu authored
The current limitation of 16 characters of the debug feature names turned out to be insufficient. Increase it to 64 characters. Signed-off-by:
-
Martin Schwidefsky authored
An arbitrary guest must not be allowed to trigger cmm actions. Only one specific guest namely the one that serves as the resource monitor may send cmm messages. Add a parameter that allows to specify the guest that may send messages. z/VMs resource manager has the name 'VMRMSVM' which is the default. Signed-off-by:
-
Peter Oberparleiter authored
Provide an easy way to define a non-zero storage key at compile time. This is useful for debugging purposes. Signed-off-by:
-
Heiko Carstens authored
The memory setup didn't take care of memory holes and this makes the memory management think there would be more memory available than there is in reality. That causes the OOM killer to kill processes even if there is enough memory left that can be written to the swap space. The patch fixes this by using free_area_init_node with an array of memory holes instead of free_area_init. Further the patch cleans up the code in setup.c by splitting setup_arch into smaller pieces. Signed-off-by:
-
Martin Schwidefsky authored
Fix overflow in calculation of the new tod value in stop_hz_timer and fix wrong virtual timer list idle time in case the virtual timer is already expired in stop_cpu_timer. Signed-off-by:
-
Martin Schwidefsky authored
Regenerate the default configuration for s390. Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Use the set_disk_ro() API when the backing file is read-only, to mark the disk read-only, during the ->open(). The current hack does not work when doing a mount -o remount. Also, mark explicitly the code paths which should no more be triggerable (I've removed the WARN_ON(1) things). They should actually become BUG()s probably but I'll avoid that since I'm not so sure the change works so well. I gave it only some limited testing. Signed-off-by:
Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> CC: Jens Axboe <axboe@suse.de> Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Add some commentary about UML internals, for a strange trick. Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Use this: .set_page_dirty = __set_page_dirty_nobuffers, We already dropped the inclusion of <linux/buffer_head.h>, and we don't have a backing block device for this FS. "Without having looked at it, I'm sure that hostfs does not use buffer_heads. So setting your ->set_page_dirty a_op to point at __set_page_dirty_nobuffers() is a reasonable thing to do - it'll provide a slight speedup." This speedup is one less spinlock held and one less conditional branch, which isn't bad. Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Fix some console locking problems (including scheduling in atomic) and various reorderings and cleanup in that code. Not yet ready for 2.6.12 probably. Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Reuse asm-x86-64/unistd.h to build our syscall table, like x86-64 already does. Like for i386, we must add some #defines for all the (right!) changes UML does to x86-64 syscall table. Note: I noted a bogus: [ __NR_sched_yield ] = (syscall_handler_t *) yield, while doing this patch (which could only be a workaround for some strange bug, but I would ignore this possibility). I'm changing this without notice. Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Fix the moved syscall table for the x86_64 SUBARCH: - redirect __NR_chown and such to versions aware of 32-bit UIDs, - avoid the useless hack for sys_nfsservctl, - use sys_sendfile64 in the table rather than sys_sendfile. - __NR_uselib is sys_ni_syscall on x86_64 (which does not support A.OUT). - __NR_getrlimit is sys_getrlimit, not sys_old_getrlimit Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Split the i386 entry.S files into entry.S and syscall_table.S which is included in the previous one (so actually there is no difference between them) and use the syscall_table.S in the UML build, instead of tracking by hand the syscall table changes (which is inherently error-prone). We must only insert the right #defines to inject the changes we need from the i386 syscall table (for instance some different function names); also, we don't implement some i386 syscalls, as ioperm(), nor some TLS-related ones (yet to provide). Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
GCC 2.95 uses __va_copy instead of va_copy. Handle it inside compiler.h instead of in a casual file, and avoid the risk that this breaks with a newer compiler (which it could do). Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
Cleanup: make an inline of this empty proc. Signed-off-by:
-
Paolo 'Blaisorblade' Giarrusso authored
We want to make possible, for the user, to enable the i586 AES implementation. This requires a restructure. - Add a CONFIG_UML_X86 to notify that we are building a UML for i386. - Rename CONFIG_64_BIT to CONFIG_64BIT as is used for all other archs - Tell crypto/Kconfig that UML_X86 is as good as X86 - Tell it that it must exclude not X86_64 but 64BIT, which will give the same results. - Tell kbuild to descend down into arch/i386/crypto/ to build what's needed. Signed-off-by:
-
Rob Landley authored
Old versions of sed from 1998 (predating the first release of gcc 2.95, but still in use by debian stable) don't understand the single-line version of the sed append command. Since newer versions of sed still understand the... ahem, "vintage" form of the command, change our code to use that. Signed-off-by:
Rob Landley <rob@landley.net> Acked-by:
Ian McDonald <imcdnzl@gmail.com> Signed-off-by:
-
Andree Leidenfrost authored
Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Fix the error path, which is triggered when the processor misses the fpx regs (i.e. the "fxsr" cpuinfo feature). For instance by VIA C3 Samuel2. Signed-off-by: -
blaisorblade@yahoo.it authored
This trick is useless, because sys_ni.c will handle this problem by itself, like it does even on UML for other syscalls. Also, it does not provide the NFSD syscall when NFSD is compiled as a module, which is a big problem. Signed-off-by:
-
Jeff Dike authored
Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Prevent the kernel from oopsing during the extable sorting, as it can do now, because the extable is in the readonly section of the binary. Jeff says: The exception table turned RO in 2.6.11-rc3-mm1 for some reason. Moving it causes it to land in the writable data section of the binary. Paolo says: This patch fixes a oops on startup, which can be easily triggered by compiling with CONFIG_MODE_TT disabled, and STATIC_LINK either disabled or enabled. The resulting kernel will always Oops on startup, after printing this simple output: I've verified, by binary search on the BitKeeper repository (synced up as of 2.6.12-rc2), starting from the range 2.6.11-2.6.12-rc1, that this bug shows up on BitKeeper revisions in the range [@1.1994.11.168,+inf), i.e. starting from this: [PATCH] lib/sort: Replace insertion sort in exception tables Since UML does not use the exception table, it's likely that insertion sort didn't happen to write anything on the table. Signed-off-by: -
Pavel Machek authored
Those cards really need A in their names. Otherwise it is pretty hard to find anything about them on the net. Signed-off-by:
Pavel Machek <pavel@suse.cz> Signed-off-by:
-
Alexander Nyberg authored
This strcpy can run off the end of saved_command_line, and we don't need it any more anyway. Signed-off-by:
Alexander Nyberg <alexn@telia.com> Cc: Andi Kleen <ak@muc.de> Signed-off-by:
-
Venkatesh Pallipadi authored
Brings sanitize_e820_map() in x86-64 in sync with that of i386. x86_64 version was missing the changes from this patch. http://linux.bkbits.net:8080/linux-2.6/cset@3e5e4083Y3HevldZl5KCy94V4DcZww?nav=index.html|src/|src/arch|src/arch/i386|src/arch/i386/kernel|related/arch/i386/kernel/setup.cSigned-off-by:
Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Acked-by:
Andi Kleen <ak@muc.de> Signed-off-by:
-
Pavel Pisa authored
Patch solves VM86 interrupt emulation deadlock on SMP systems. The VM86 interrupt emulation has been heavily tested and works well on UP systems after last update, but it seems to deadlock when we have used it on SMP/HT boxes now. It seems, that disable_irq() cannot be called from interrupts, because it waits until disabled interrupt handler finishes (/kernel/irq/manage.c:synchronize_irq():while(IRQ_INPROGRESS);). This blocks one CPU after another. Solved by use disable_irq_nosync. There is the second problem. If IRQ source is fast, it is possible, that interrupt is sometimes processed and re-enabled by the second CPU, before it is disabled by the first one, but negative IRQ disable depths are not allowed. The spinlocking and disabling IRQs over call to disable_irq_nosync/enable_irq is the only solution found reliable till now. Signed-off-by:
Michal Sojka <sojkam1@control.felk.cvut.cz> Signed-off-by:
Pavel Pisa <pisa@cmp.felk.cvut.cz> Signed-off-by:
-
Venkatesh Pallipadi authored
The specifications that talk about E820 map doesn't have an upper limit on the number of e820 entries. But, today's kernel has a hard limit of 32. With increase in memory size, we are seeing the number of E820 entries reaching close to 32. Patch below bumps the number upto 128. The patch changes the location of EDDBUF in zero-page (as it comes after E820). As, EDDBUF is not used by boot loaders, this patch should not have any effect on bootloader-setup code interface. Patch covers both i386 and x86-64. Tested on: * grub booting bzImage * lilo booting bzImage with EDID info enabled * pxeboot of bzImage Side-effect: bss increases by ~ 2K and init.data increases by ~7.5K on all systems, due to increase in size of static arrays. Signed-off-by:
-
