- 20 Oct, 2004 29 commits
-
-
Hugh Dickins authored
Let's lighten the global spinlock mmlist_lock. What's it for? 1. Its original role is to guard mmlist. 2. It later got a second role, to prevent get_task_mm from raising mm_users from the dead, just after it went down to 0. Firstly consider the second: __exit_mm sets tsk->mm NULL while holding task_lock before calling mmput; so mmlist_lock only guards against the exceptional case, of get_task_mm on a kernel workthread which did AIO's use_mm (which transiently sets its tsk->mm without raising mm_users) on an mm now exiting. Well, I don't think get_task_mm should succeed at all on use_mm tasks. It's mainly used by /proc/pid and ptrace, seems at best confusing for those to present the kernel thread as having a user mm, which it won't have a moment later. Define PF_BORROWED_MM, set in use_mm, clear in unuse_mm (though we could just leave it), get_task_mm give NULL if set. Secondly consider the first: and what's mmlist for? 1. Its original role was for swap_out to scan: rmap ended that in 2.5.27. 2. In 2.4.10 it got a second role, for try_to_unuse to scan for swapoff. So, make mmlist a list of mms which maybe have pages on swap: add mm to mmlist when first swap entry is assigned in try_to_unmap_one (pageout), or in copy_page_range (fork); and mmput remove it from mmlist as before, except usually list_empty and there's no need to lock. drain_mmlist added to swapoff, to empty out the mmlist if no swap is then in use. mmput leave mm on mmlist until after its exit_mmap, so try_to_unmap_one can still add mm to mmlist without worrying about the mm_users 0 case; but try_to_unuse must avoid the mm_users 0 case (when an mm might be removed from mmlist, and freed, while it's down in unuse_process): use atomic_inc_return now all architectures support that. Some of the detailed comments in try_to_unuse have grown out of date: updated and trimmed some, but leave SWAP_MAP_MAX for another occasion. Signed-off-by:
Hugh Dickins <hugh@veritas.com> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Hirofumi Ogawa authored
The low byte of FAT's first entry must have same value with media-field. So, I added that validity check for some case. In fact, in some case this check detected non-FAT filesystem. But in real world, unfortunately too many devices is writing a wrong value. This patch removes that validity check. Signed-off-by:
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> Signed-off-by:
-
Hirofumi Ogawa authored
Signed-off-by:
-
Hirofumi Ogawa authored
This patch make sure it's <= sbi->clusters. Signed-off-by:
-
Hirofumi Ogawa authored
Signed-off-by:
-
Hirofumi Ogawa authored
Signed-off-by:
-
Hirofumi Ogawa authored
This patch fixes the race condition | fat_free | fat_get_cluster +------------------------------+----------------------- fat_cache_lookup() (get the copy of cache) fat_cache_inval_inode() (invalidate caches on inode) fat_cache_add() (update/add the getted cache) The above race has possible that invalidated cache is added. This patch fixes the race condition by adding the cache-id to copy of cache. Signed-off-by: -
Hirofumi Ogawa authored
Signed-off-by:
-
Hirofumi Ogawa authored
Signed-off-by:
-
Hirofumi Ogawa authored
This rewrites the cache stuff for file allocation table (FAT). This cache stocks the more pieces of FAT-chain by counting the number of contiguous data blocks. And if cache hit, since a block number can calculate without looking FAT-chain up, fat_get_block() become more fast. But if data blocks was fragmenting, unfortunately this cache is unuseful. read from block device [1st]# time dd if=/dev/hda6 of=/dev/null bs=1M count=2048 2147483648 bytes transferred in 229.524189 seconds (9356241 bytes/sec) real 3m49.557s, user 0m0.026s, sys 1m20.414s [2nd]# time dd if=/dev/hda6 of=/dev/null bs=1M count=2048 2147483648 bytes transferred in 229.539358 seconds (9355623 bytes/sec) real 3m49.647s, user 0m0.036s, sys 1m20.144s read from full contiguous file with this patch [1st]# time dd if=data of=/dev/null bs=1M count=2048 2147483648 bytes transferred in 337.959477 seconds (6354264 bytes/sec) real 5m37.970s, user 0m0.031s, sys 1m21.915s [2nd]# time dd if=data of=/dev/null bs=1M count=2048 2147483648 bytes transferred in 225.401699 seconds (9527362 bytes/sec) real 3m45.476s, user 0m0.027s, sys 1m19.286s read from full fragmented file with this patch [1st]# time dd if=data of=/dev/null bs=1M count=2048 2147483647 bytes transferred in 1146.529081 seconds (1873030 bytes/sec) real 19m6.538s, user 0m0.020s, sys 1m32.774s [2nd]# time dd if=data of=/dev/null bs=1M count=2048 2147483647 bytes transferred in 1045.084822 seconds (2054841 bytes/sec) real 17m25.152s, user 0m0.022s, sys 1m34.801s read from full contiguous file without this patch [1st]# time dd if=data of=/dev/null bs=1M count=2048 2147483648 bytes transferred in 348.144707 seconds (6168365 bytes/sec) real 5m48.169s, user 0m0.019s, sys 1m29.962s [2nd]# time dd if=data of=/dev/null bs=1M count=2048 2147483648 bytes transferred in 324.017361 seconds (6627681 bytes/sec) real 5m24.038s, user 0m0.023s, sys 1m20.602s read from full fragmented file without this patch [1st]# time dd if=data of=/dev/null bs=1M count=2048 2147483647 bytes transferred in 1156.845693 seconds (1856327 bytes/sec) real 19m16.855s, user 0m0.031s, sys 1m32.172s [2nd]# time dd if=data of=/dev/null bs=1M count=2048 2147483647 bytes transferred in 1066.518713 seconds (2013545 bytes/sec) real 17m46.526s, user 0m0.023s, sys 1m33.630s Signed-off-by:
-
Hirofumi Ogawa authored
Signed-off-by:
-
Christoph Hellwig authored
These were used by the defunct umsdos code only. Signed-off-by:
-
William Lee Irwin III authored
Add atomic_sub_and_test() to sparc32, implemented in terms of atomic_sub_return(), so reiser4 can be simultaneously microoptimized for x86 and made to pass compilation testing on sparc32. Signed-off-by:
-
KaiGai Kohei authored
This patch declares atomic_add_return() as an alias of __atomic_add(). atomic64_add_return(),atomic_sub_return() and atomic64_sub_return() are same. Signed-off-by:
KaiGai, Kohei <kaigai@ak.jp.nec.com> Signed-off-by:
-
KaiGai Kohei authored
This patch implements atomic_inc_return() and so on for ARM26. Because Hugh says that SMP is not supported in arm26, it is implemented by normal operations between local_irq_save() and local_irq_restore() like another atomic operations. Signed-off-by:
-
KaiGai Kohei authored
This patch declares atomic_inc_return() as the alias of atomic_add_return() and atomic_dec_return() as an alias of atomic_dec_return(). Signed-off-by:
-
KaiGai Kohei authored
Signed-off-by:
-
KaiGai Kohei authored
This patch implements atomic_inc_return() and so on for i386, and includes runtime check whether CPU is legacy 386. Signed-off-by:
-
David Woodhouse authored
We didn't use the 'usercompr' field before recently. Unfortunately we didn't set it to zero either, so there's a lot of file systems out there with it set to 0x5a. Deal with that rather than getting confused. Signed-Off-By:
David Woodhouse <dwmw2@infradead.org> Signed-off-by:
-
Randy Dunlap authored
Code section errors in i386/io_apic.c found by scripts/reference_init.pl. Looks like they could cause problems for a few drivers or in a real hotplug environment. Error: ./arch/i386/kernel/io_apic.o .text refers to 000018ff R_386_PC32 .init.text call chain: snd_mpu401_acpi_resource acpi_register_gsi mp_register_gsi io_apic_set_pci_routing {A} ioapic_register_intr IO_APIC_irq_trigger find_irq_entry Error: ./arch/i386/kernel/io_apic.o .text refers to 00001967 R_386_PC32 .init.text (as above thru {A}, then:) IO_APIC_irq_trigger irq_trigger MPBIOS_trigger >> removing __init from this led to needing to remove __init from EISA_ELCR also. Signed-off-by:Randy Dunlap <rddunlap@osdl.org> Signed-off-by:
-
Gordon Jin authored
Fix a bug that arch/x86_64/lib/copy_user:copy_user_generic will return a wrong value when exception happens. In the case the address is not 8-byte aligned (i.e. go into Lbad_alignment), if exception happens in Ls11, %rdx will be wrong number of copied bytes, then copy_user_generic returns wrong value. It also fixed a bug of zeroing wrong number of bytes of destination at this situation. (In Lzero_rest) Signed-off-by:
Yanmin Zhang <yanmin.zhang@intel.com> Signed-off-by:
Nanhai Zou <nanhai.zou@intel.com> Signed-off-by:
Gordon Jin <gordon.jin@intel.com> Signed-off-by:
Suresh Siddha <suresh.b.siddha@intel.com> Signed-off-by:
-
Andi Kleen authored
Split CONFIG_NUMA_EMU and CONFIG_K8_NUMA because they are independent. Signed-off-by:
Andi Kleen <ak@muc.de> Signed-off-by:
-
Andi Kleen authored
Originally suggested by Zwane Mwaikumbo Ignore all ACPI timer overrides on all Nvidia boards. The fallback doesn't work and no Nvidia boards needs a timer override. But some buggy BIOS have it anyways. Also enable IO-APIC mode by default for Nvidia then. Thanks to Andy Currid for confirming this. Signed-off-by:
Andi Kleen <ak@suse.de> Signed-off-by:
-
Andi Kleen authored
Add an kstack= option to configure how much stack should be printed on a oops. Signed-off-by:
-
Andi Kleen authored
Add an notsc option so that it can be turned off again. This may be useful on the Summit, but will only work when there is a HPET fallback. Signed-off-by:
-
Andi Kleen authored
x86-64 MP kernels always used HPET timing when available. This was done because AMD systems don't have synchronized TSCs in all cases. On Intel this is not true (except Summit), so use faster TSC timing for those. Also only enable HPET gettimeofday when there is more than one CPU online. Signed-off-by:
-
Andi Kleen authored
Initialize HPET char driver Signed-off-by:
-
Andi Kleen authored
Suggested by Ingo Molnar. Drop APIC level workaround for old IO-APICs on x86-64 because it is a significant part of interrupt handling time. x86-64 only runs on modern chipsets that probably don't have this bug. Signed-off-by:
-
Rusty Russell authored
module_param_array() takes a variable to put the number of elements in. Looking through the uses, many people don't care, so they declare a dummy or share one variable between several parameters. The latter is problematic because sysfs uses that number to decide how many to display. The solution is to change the variable arg to a pointer, and if the pointer is NULL, use the "max" value. This change is fairly small, but fixing up the callers is a lot of (trivial) churn. Signed-off-by:
Rusty Russell <rusty@rustcorp.com.au> Signed-off-by:
-
- 19 Oct, 2004 9 commits
-
-
http://lia64.bkbits.net/linux-ia64-release-2.6.10Linus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
Tony Luck authored
into agluck-lia64.sc.intel.com:/data/home/aegl/BK/linux-ia64-release-2.6.10
-
Pat Gefre authored
Signed-off-by:
Patrick Gefre <pfg@sgi.com> Signed-off-by:
Tony Luck <tony.luck@intel.com>
-
Robin Holt authored
During peak utilization periods, the first interface on a node has an inordinately large amount of contention. This is due to all cpus starting their scan for an interface at 0. This patch distributes that based upon the slice the requesting cpu is attached to. Signed-off-by: Robin Holt Signed-off-by: -
Robin Holt authored
The SN2 Block Transfer Engine occassionally fails to send a notification that it has completed a transfer to the kernel. This patch adds a timeout mechanism which will detect the failure, reset the interface, and then retry the transfer. Signed-off-by: Robin Holt Signed-off-by: -
Pat Gefre authored
Signed-off-by:
-
Robin Holt authored
If all bte interfaces are in use, the current code will leave the bte pointer set when exiting from the loop trying to locate an interface to use. This results in two processes using the same interface and both trying to free the same one. With a premptible kernel, this results in the preempt count getting off. Signed-off-by:
Robin Holt <holt@sgi.com> Signed-off-by:
-
bk://kernel.bkbits.net/gregkh/linux/i2c-2.6Linus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
bk://bk.arm.linux.org.uk/linux-2.6-serialLinus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
- 20 Oct, 2004 1 commit
-
-
Russell King authored
This includes a fuller definition of the 8250 and compatible bits, and adding notes where the definition varies between different chips. This also fixes the following build error: drivers/serial/8250.c:185: error: `UART_FCR_R_TRIG_10'
-
- 19 Oct, 2004 1 commit
-
-
Russell King authored
Some serial drivers receive their serial port device information via the device model. This unfortunately means that the selected port may not be available when the console subsystem initialises, so we must keep trying to register the console after each port is added.
-
