Andrew Morton [Thu, 19 Feb 2004 00:40:04 +0000 (16:40 -0800)]
[PATCH] show_task() is not SMP safe
From: Arnd Bergmann <arnd@arndb.de>
Christian Bornträger noticed that the kernel can crash after <SysRq>-T. It
appears that the show_task function gets called for all tasks, which does
not work if one of the tasks is running in a system call on another CPU.
In that case the result of thread_saved_pc and show_stack is undefined and
likely to cause a crash.
For tasks running in user space on other CPUs, show_task() is probably
harmless, but I'm not sure if that's true on all architectures.
The patch below is still racy for tasks that are about to sleep, but it
demonstrates the problem.
Andrew Morton [Thu, 19 Feb 2004 00:39:53 +0000 (16:39 -0800)]
[PATCH] show_task() fix and cleanup
show_task() is preinting negative numbers for free stack due to arithmetic
against the wrong pointer.
Fix that up, and clean up a few related things.
show_task still has bogus code which atempts to work out how much stack the
task has ever used - it cannot work because we don't actually zero out the
stack pages when allocating them. We should fix that, or take it out.
Andrew Morton [Thu, 19 Feb 2004 00:39:43 +0000 (16:39 -0800)]
[PATCH] print some x86 build options during oopses
I find this handy sometimes: it makes the oops output include info as to
whether the user has selected CONFIG_PREEMPT, CONFIG_SMP or, particularly,
CONFIG_DEBUG_PAGEALLOC. It can save one email round-trip.
Andrew Morton [Thu, 19 Feb 2004 00:39:33 +0000 (16:39 -0800)]
[PATCH] config option for irqbalance
From: "Martin J. Bligh" <mbligh@aracnet.com>
Make irqbalance into a config option - some people (jgarzik, arjan, etc)
wanted to be able to disable it and do things from userspace instead. This
patch allows each camp to do their own thing, which seems fair ;-)
Andrew Morton [Thu, 19 Feb 2004 00:39:12 +0000 (16:39 -0800)]
[PATCH] ia32: pfn_to_nid fix
From: "Martin J. Bligh" <mbligh@aracnet.com>
Makes sure pfn_to_nid is defined for all combinations of subarches, and that
it's defined before it's used so we don't run into implicit declaration
problems.
Andrew Morton [Thu, 19 Feb 2004 00:39:02 +0000 (16:39 -0800)]
[PATCH] fix pfn_valid on ia32 discontigmem
From: "Martin J. Bligh" <mbligh@aracnet.com>
Fix pfn_valid for architctures with discontiguous memory. This only
changes the NUMA definition, and it leaves the NUMA-Q definition as was,
because it's faster that way, it's in hotpaths, and our memory is always
contiguous.
Andrew Morton [Thu, 19 Feb 2004 00:38:51 +0000 (16:38 -0800)]
[PATCH] fix access() POSIX compliance
From: Andreas Gruenbacher <agruen@suse.de>
The fix for permission() that makes it compliant with POSIX.1-2001
apparently was lost. Here is the patch I sent before. (The relevant lines
from the standard text are cited in
http://www.ussg.iu.edu/hypermail/linux/kernel/0310.2/0286.html. The fix
proposed in that posting did not handle directories without execute
permissions correctly.)
Make permission check conform to POSIX.1-2001
The access(2) function does not conform to POSIX.1-2001: For root
and a file with no permissions, access(file, MAY_READ|MAY_EXEC)
returns 0 (it should return -1).
Andrew Morton [Thu, 19 Feb 2004 00:38:29 +0000 (16:38 -0800)]
[PATCH] use noinline for rest_init()
gcc-3.4 incorretly inlines rest_init() into start_kernel(), causing things to
crash when the .text.init section gets unloaded. Use noinline to prevent
that.
Andrew Morton [Thu, 19 Feb 2004 00:38:18 +0000 (16:38 -0800)]
[PATCH] Add noinline attribute
From: Andi Kleen <ak@muc.de>
This patch adds the `noinline' function attribute. It can be used to
explicitly tell the compiler to not inline functions.
We need this due to what is, IMO, a bug present in gcc-3.4 and current
gcc-3.5 CVS: the compiler is inlining init/main.c:rest_init() inside
init/main.c:start_kernel(), despite the fact that thay are declared to be
placed in different text sections.
Andrew Morton [Thu, 19 Feb 2004 00:38:07 +0000 (16:38 -0800)]
[PATCH] Use -funit-at-a-time on ia32
From: Andi Kleen <ak@muc.de>
The upcomming gcc 3.4 has a new compilation mode called unit-at-a-time.
What it does is to first load the whole file into memory and then generate
the output. This allows it to use a better inlining strategy, drop unused
static functions and use -mregparm automatically for static functions.
It does not seem to compile significantly slower.
This is also available in some of the 3.3 based "hammer branch"
compilers used in distributions (at least in SuSE and Mandrake)
Some tests show impressive .text shrinkage from unit-at-a-time.
e.g. here is the same kernel compiled with -fno-unit-at-a-time and
-funit-at-a-time with a gcc 3.4 snapshot. The gains are really
impressive:
text data bss dec hex filename 4129346 708629 207240 5045215 4cfbdf vmlinux-nounitatatime 3999250 674853 207208 4881311 4a7b9f vmlinux-unitatatime
.text shrinks by over 130KB!. And .data shrinks too.
At first look the numbers look nearly too good to be true, but they have been
verified with several configurations and seem to be real. It looks like
we have a lot of stupid inlines or dead functions. I'm really not
sure why it is that much better. But it's hard to argue with hard
numbers.
[A bloat-o-meter comparision between the two vmlinuxes can be found in
http://www.firstfloor.org/~andi/unit-vs-no-unit.gz . It doesn't show
any obvious candidates unfortunately, just lots of small changes]
With the gcc 3.3-hammer from SuSE 9.0 the gains are a bit smaller, but
still noticeable (>100KB on .text)
This patch enables -funit-at-a-time on ia32 if the compiler is gcc-3.4 or
later. We had several reports of gcc-3.3 producing very early lockups.
This one helps even more, >236KB .text difference. Clearly worth
the effort.
This patch adds an option to use -mregparm=3 while compiling the kernel. I
did an LTP run and it showed no additional failures over an non regparm
kernel.
According to some gcc developers it should be safe to use in all gccs that
are still supports (2.95 and up)
I didn't make it the default because it will break all binary only modules
(although they can be fixed by adding a wrapper that calls them with
"asmlinkage"). Actually it may be a good idea to make this default with
2.7.1 or somesuch.
We add new kbuild infrastructure: the command
scripts/gcc-version.sh $(CC)
will print out the version of gcc in a canonical 4-digit form suitable for
performing numerical tests against.
DESC
arch/i386/Makefile,scripts/gcc-version.sh,Makefile small fixes
EDESC
From: Serge Belyshev <33554432@mtu-net.ru>
arch/i386/Makefile:
* omitted $(KBUILD_SRC)/ in script call.
scripts/gcc-version.sh:
* GNU tail no longer supports 'tail -1' syntax.
We should consider adding -fweb option:
vanilla:
$ size vmlinux
text data bss dec hex filename 3056270 526780 386056 3969106 3c9052 vmlinux
with -fweb:
$ size vmlinux
text data bss dec hex filename 3049523 526780 386056 3962359 3c75f7 vmlinux
Also note 0.1 ... 1.0% speedup in various benchmarks.
This option is not enabled by default at -O2 because it
(like -fomit-frame-pointer) makes debugging impossible.
Andrew Morton [Thu, 19 Feb 2004 00:37:32 +0000 (16:37 -0800)]
[PATCH] slab: hexdump for check_poison
From: Manfred Spraul <manfred@colorfullife.com>
The patch is designed improve the diagnostics which are presented when the
slab memory poison detector triggers.
check_poison_obj checks for write accesses after kfree by comparing the
object contents with the poison value. The current implementation contains
several flaws:
- it accepts both POISON_BEFORE and POISON_AFTER. check_poison_obj is
only called with POISON_AFTER poison bytes. Fix: only accept
POISON_AFTER.
- the output is unreadable. Fix: use hexdump.
- if a large objects is corrupted, then the relevant lines can scroll of
the screen/dmesg buffer. Fix: line limit.
- it can access addresses behind the end of the object, which can oops
with CONFIG_DEBUG_PAGEALLOC. Fix: bounds checks.
Additionally, the patch contains the following changes:
- rename POISON_BEFORE and POISON_AFTER to POISON_FREE and POISON_INUSE.
The old names are ambiguous.
- use the new hexdump object function in ptrinfo.
- store_stackinfo was called with wrong parameters: it should store
caller, i.e. __builtin_return_address(0), not POISON_AFTER in the
object.
- dump both the object before and after the corrupted one, not just the
one after.
Andrew Morton [Thu, 19 Feb 2004 00:37:22 +0000 (16:37 -0800)]
[PATCH] prevent ptrace from altering page permissions
From: Roland McGrath <roland@redhat.com>
Under some circumstances, ptrace PEEK/POKE_TEXT can cause page permissions
to be permanently changed. Thsi causes changes in application behaviour
when run under gdb.
Fix that by only marking the pte as writeable if the vma is marked for
writing. A write fault thus unshares the page but doesn't necessarily make
it writeable.
Andrew Morton [Thu, 19 Feb 2004 00:37:00 +0000 (16:37 -0800)]
[PATCH] ramdisk cleanup
Fairly pointless coding-style cleanups which I've been sitting on for ages.
The ramdisk driver is still buggy: it drops pagecache when unmounted. I
still need to fix this.
Apparently it also displays data corruption under load even when not
unmounted.
Greg Ungerer [Wed, 18 Feb 2004 14:15:10 +0000 (06:15 -0800)]
[PATCH] fixes to ColdFire/5407 startup code
Correct the cache setup bits for the 5407. This enables the write
buffers properly (despite what the previous comment said). This
combined with fixed cache flushing code provides a nice performance
boost on the 5407.
Also fix the ROMfs setup to only move the ROMfs region if it is actually
configured.
Greg Ungerer [Wed, 18 Feb 2004 13:41:28 +0000 (05:41 -0800)]
[PATCH] allow configuration for shared flat binary support
This adds the configuration option to enable the uClinux shared flat
binary support. The code support is already in the binfmt_load code,
just the config option is missing.
Andrew Morton [Wed, 18 Feb 2004 13:00:09 +0000 (05:00 -0800)]
[PATCH] mark ftape un-removable
From: Christoph Hellwig <hch@lst.de>
I've just grepped over the tree for reamining MOD_INC_USE_COUNT users, and
ftape is a really horrible one, it relies on MOD_{INC,DEV}_USE_COUNT in the
most horrible places + it's own bookkepping and the <= 2.4 may unload
hooks. And although I don't have the hardware I can guarantee it doesn't
work as expected.
So let's just remove the module_exit handler and mark it unremovable, if
someone wants to fix up ftape later it's fine with me, but it's a really
scary driver..
Andrew Morton [Wed, 18 Feb 2004 12:59:51 +0000 (04:59 -0800)]
[PATCH] crc32.c copyright fix
From: Matt Domsch <Matt_Domsch@dell.com>
Patch below applies to both 2.4.25 and 2.6.3, and replaces the public
domain statement and non-warranty with the GPL, as is permitted by the code
being in the public domain, and is done with legal advice.
Andrew Morton [Wed, 18 Feb 2004 12:59:42 +0000 (04:59 -0800)]
[PATCH] tuner driver fixes
From: Gerd Knorr <kraxel@bytesex.org>
"options tuner type=2" is just there for historical reasons and should
only be used/needed if the main driver doesn't allow to configure the
tuner type. I'm not sure whenever I can remove that altogether without
breaking anything. There are several users of the tuner module, some
outside the standard kernel tree ...
> > if (type < TUNERS) {
> > + t->type = type;
> > printk("tuner: type forced to %d (%s) [insmod]\n",
> > t->type,tuners[t->type].name);
> > set_type(client,type);
That is wrong, it will break in other corner cases, it may cause the
set_type() function not doing the initializations.
The prink can also be dropped because set_type does that too. The patch
below removes that. It also makes the kernel messages a bit more verbose
and adds support for two new tuners.
The file initramfs_data.cpio is slightly different when generated on
cygwin, compared to linux, which causes the kernel to panic with the
message "no cpio magic" (See Documentation/early-userspace/README).
The problem in cpio generation is due to the difference in sprintf
modifiers on cygwin. The code uses "%08ZX" for strlen of a device node.
printf man pages discourages "Z" and has 'z' instead. Both of these are
not available on cygwin sprintf (at least some versions of cygwin). The
net result of all of this is that the generated file literally contains
"ZX" and then the strlen after that and messes up that 110 offset etc. The
file is 516 bytes long on the system that I tested and on linux it is 512
bytes.
Andrew Morton [Wed, 18 Feb 2004 12:58:57 +0000 (04:58 -0800)]
[PATCH] Remove overenthusiastic BUG in smp_boot_cpus
From: "Martin J. Bligh" <mbligh@aracnet.com>
There's no real need to BUG and stop the system from booting if the BIOS
spec'ed apicid for the boot CPU isn't correct - we can continue perfectly
well with the real one.
I created the little wrapper function that was requested. I think the only
other arch, other than ia64, that doesn't at least include asm-generic/tlb.h
is arm.
Something appears broken in TLB flushing on IA64 (& possibly other
architectures). Functionally, it works but performance is bad on systems
with large cpu counts.
The result is that TLB flushing in exit_mmap() is frequently being done via
IPIs to all cpus rather than with a "ptc" instruction or with a new
context..
Andrew Morton [Wed, 18 Feb 2004 12:58:20 +0000 (04:58 -0800)]
[PATCH] Fix make rpm when using RH9 or Fedora..
From: Thomas Davis <tadavis@lbl.gov>
Doing a 'make rpm' will fail with the current RH9/Fedora RPM macros.
The failure message is this:
Processing files: kernel-debuginfo-2.6.3rc1mm1-12
error: Could not open %files file /usr/src/redhat/BUILD/kernel-2.6.3rc1mm1/debugfiles.list: No such file or directory
Andrew Morton [Wed, 18 Feb 2004 12:58:11 +0000 (04:58 -0800)]
[PATCH] dm-crypt
From: Christophe Saout <christophe@saout.de>
Adds a crypto module for device-mapper. The intent here is to remove
cryptoloop ASAP, to pull the remapping gunk out of the loops driver and to
migrate people onto dm-crypt. It is on-disk compatible with existing
cryptolop installations.
See http://www.saout.de/misc/dm-crypt/ for usage details.
Andrew Morton [Wed, 18 Feb 2004 12:57:52 +0000 (04:57 -0800)]
[PATCH] ide-tape: remove obsolete onstream support
From: Willem Riede <wrlk@riede.org>
The onstream drives, be they scsi, atapi, ieee1394 or usb, are supported by the
osst driver. When the drives were new, code was introduced in ide-tape
independently, but never really maintained since. There are issues with the
drives I found and dealt with in osst, that ide-tape doesn't address.
So this code in ide-tape is both redundant and imperfect. I assumed from
http://marc.theaimsgroup.com/?l=linux-kernel&m=107550547613846&w=2 that there
was buy in for removing it.
When this patch is applied, ide-tape will refuse to attach to an onstream DI
drive, and will write to syslog to use ide-scsi + osst instead.
Andrew Morton [Wed, 18 Feb 2004 12:57:25 +0000 (04:57 -0800)]
[PATCH] Fix race in epoll_ctl(EPOLL_CTL_MOD)
From: Davide Libenzi <davidel@xmailserver.org>
A potential race can happen in epoll_ctl(EPOLL_CTL_MOD) where an event can
happen in between f_op->poll() and the lock on ep->lock (we cannot call
f_op->poll() inside a lock, and the f_op->poll() callback does not carry
any info at the current time - missing wake_up_info() already ;). In that
case the event would be removed. We can easily leave the event inside the
ready list and have the ep_send_events() logic do the job for us at later
time. (Thanks to david.lee@teracruz.com for reporting the thing, since it
shouldn't have been a nice one ;)
Andrew Morton [Wed, 18 Feb 2004 12:56:48 +0000 (04:56 -0800)]
[PATCH] Fix __release_region() race
From: MAEDA Naoaki <maeda.naoaki@jp.fujitsu.com>
I am testing PCI hot-plug in 2.6.2 kernel, but sometimes a struct resource
tree in kernel/resource.c was broken if multiple hot-plug requests are
issued at the same time.
The reason is lots of drivers call release_region() on hot removal, and
__release_region(), which is invoked by release_region() macro, changes the
tree without holding a writer lock for resource_lock.
I think __release_region() must hold a writer lock as well as
__request_region() does.
A following patch fixes the issue in my environment.
Andrew Morton [Wed, 18 Feb 2004 12:56:30 +0000 (04:56 -0800)]
[PATCH] oprofile: add Pentium Mobile support
From: Philippe Elie <phil.el@wanadoo.fr>
From: Will Cohen <wcohen@redhat.com>
Add oprofile support for Pentium Mobile (P6 core). Pentium Mobile needs to
unmask LVPTC vector, since it doesn't hurt other P6 core based cpus we do
it unconditionally for all these.
This patch require userspace tools >= 0.8 (only in sourceforge cvs currently)
Andrew Morton [Wed, 18 Feb 2004 12:56:21 +0000 (04:56 -0800)]
[PATCH] oprofile: ARM infrastructure
From: Philippe Elie <phil.el@wanadoo.fr>
From: Zwane Mwaikambo <zwane@arm.linux.org.uk>
This patch adds infrastructure code and enables ARM to utilise the timer
int oprofile driver. There is PMU code under development for the XScale
but that is still forthcoming. In the meantime you can use the timer int
driver with an updated Oprofile-CVS userspace (SF is a bit slow, please
allow 24hrs).
Andrew Morton [Wed, 18 Feb 2004 12:55:53 +0000 (04:55 -0800)]
[PATCH] cleanup patch that prepares for 4Kb stacks
From: Arjan van de Ven <arjanv@redhat.com>
I have 4Kb stacks + IRQ stacks working in my tree. The biggest part of the
4K-stacks work is changing hardcoded 8Kb assumptions to the proper,
pre-existing define for this. That part of the patch is appropriate in
general, even when 4Kb stacks might not be.
Andrew Morton [Wed, 18 Feb 2004 12:55:44 +0000 (04:55 -0800)]
[PATCH] fix display of NBD in /proc/partitions
The recent change to /proc/partitions which prevents it from displaying
removeable media accidentally caused 128 NBD and 16 ramdisk partitions to
appear in /proc/partitions instead.
So add a specific gendisk flag which says "don't show me in /proc/partitions"
and use that.
Andrew Morton [Wed, 18 Feb 2004 12:55:26 +0000 (04:55 -0800)]
[PATCH] off_t in nfsd_commit needs to be loff_t
From: Neil Brown <neilb@cse.unsw.edu.au>,
From: Miquel van Smoorenburg <miquels@cistron.nl>
While I was stress-testing NFS/XFS on 2.6.1/2.6.2-rc, I found that
sometimes my "dd" would exit with:
# dd if=/dev/zero bs=4096 > /mnt/file
dd: writing `standard output': Invalid argument 1100753+0 records in 1100752+0 records out
After adding some debug printk's to the server and client code and some
tcpdump-ing, I found that the NFSERR_INVAL was returned by nfsd_commit on
the server.
Turns out that the "offset" argument is off_t instead of loff_t. It isn't
used at all (unfortunately), but it _is_ checked for sanity, so that's
where the error came from.
Andrew Morton [Wed, 18 Feb 2004 12:55:17 +0000 (04:55 -0800)]
[PATCH] drivers/char/vt possible race
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
I falled again on the crash in con_do_write() with driver->data beeing
NULL. It happens during boot, when userland is playing open/close games
with tty's, I was intentionally typing keys like mad during boot trying to
trigger another problem when this one poped up.
Looking at the code, I'm not sure how protected we are by the above (tty)
layer, paulus told me to not rely on anything like locking coming from
there, so I decided to extend the scope of the console semaphore one more
bit to cover races between calls to con_open, con_close and con_write.
Note that in con_do_write, I intentionally drop the semaphore to avoid
keeping it held when waiting on the local buffer, and I added some sanity
checks on tty->driver_data with some printk's in case we still have an open
race by the tty layer. At least, now, the couple vc_allocated &
tty->driver_data should be protected though.
I had a look at fs/proc/base.c and found that the 'lookup' functions for
these directories were checking that the task in question existed, but
overlooked the following:
1. In the function proc_pid_lookup, a check is required to ensure that
the task in question is a thread group leader. Without the check, any
task can have its data retrieved accordingly. Consider the following.
There is a multithreaded process 1777.
2. The other part of the bug is in the function proc_task_lookup. Here
there needs to be a check that the task X is indeed a thread of the
thread group Y when we read /proc/<Y>/task/<X>.
Right now, this check does not exist, which allows for any existing
task to have its data read from another thread group directory. The
following reads the stat directory of my bash shell from the thread
group 1.
gen2 23:28:07 ~: cd /proc/1
gen2 23:28:10 /proc/1: ls
auxv cwd exe maps mounts stat status wchan
cmdline environ fd mem root statm task
gen2 23:28:11 /proc/1: ls task
1
gen2 23:28:27 /proc/1: cat task/$$/stat
1671 (bash) S 1670 1671 1671 34817 1802 256 1953 8101 12 4 10 6 9 26 15 0 1 0 5789 3252224 454 4294967295134512640134955932322122510432212228404294960144 0 65536 368640412667614673222442959 0 0 17 0 0 0
Andrew Morton [Wed, 18 Feb 2004 12:54:41 +0000 (04:54 -0800)]
[PATCH] defer panic for too many items in boot parameter line
From: Werner Almesberger <werner@almesberger.net>
When passing too many unrecognized boot command line options (which become
arguments or environment variables), the 2.6 kernel panics (unlike 2.4,
which just ignores the extra items). Unfortunately, this happens before
the console is initialized, so all you get is a kernel that dies quickly,
for no apparent reason.
This is particularly irritating if using UML with
init=something wi th a lot of ar gu men t s
The patch below delays the panic until after console_init.
(akpm: I mainly added this in because we have other places where the
panic-later-on machinery is needed).
Andrew Morton [Wed, 18 Feb 2004 12:53:55 +0000 (04:53 -0800)]
[PATCH] Rename bitmap_snprintf() and cpumask_snprintf() to *_scnprintf()
From: Joe Korty <joe.korty@ccur.com>
Rename bitmap_snprintf() to bitmap_scnprintf() and cpumask_snprintf() to
cpumask_scnprintf(), as these functions now belong to the scnprintf family
of functions.
Andrew Morton [Wed, 18 Feb 2004 12:53:28 +0000 (04:53 -0800)]
[PATCH] 8259 timer ack fix
From: "Maciej W. Rozycki" <macro@ds2.pg.gda.pl>
Fix up the 8259 ack handling for buggy SMM firmware.
See http://www.ussg.iu.edu/hypermail/linux/kernel/0203.2/0956.html
Apparently the embedded 8259A-compatible core is not fully functional.
This patch lets the I/O APIC-driven NMI watchdog to function correctly.
Credit to Ross Dickson for discovering this.
Andrew Morton [Wed, 18 Feb 2004 12:53:19 +0000 (04:53 -0800)]
[PATCH] dm: drop BIO_SEG_VALID bit
From: Joe Thornber <thornber@redhat.com>
I just noticed that bio_clone copies the BIO_SEG_VALID bit from the original
bio when it was set. When we modify bi_idx or bi_vcnt afterwards the segment
counts are invalid and the bit must be dropped (though it is fairly unlikely
that it has already been set). [Christophe Saout]
Andrew Morton [Wed, 18 Feb 2004 12:51:58 +0000 (04:51 -0800)]
[PATCH] md: Allow partitioning of MD devices.
From: NeilBrown <neilb@cse.unsw.edu.au>
With this patch, md used two major numbers for arrays.
One Major is number 9 with name 'md' have unpartitioned md arrays, one per
minor number.
The other Major is allocated dynamically with name 'mdp' and had on array for
every 64 minors, allowing for upto 63 partitions.
The arrays under one major are completely separate from the arrays under the
other.
The preferred name for devices with the new major are of the form:
/dev/md/d1p3 # partion 3 of device 1 - minor 67
When a paritioned md device is assembled, the partitions are not recognised
until after the whole-array device is opened again. A future version of
mdadm will perform this open so that the need will be transparent.
Andrew Morton [Wed, 18 Feb 2004 12:51:49 +0000 (04:51 -0800)]
[PATCH] md: Dynamically limit size of bio requests used for raid1 resync
From: NeilBrown <neilb@cse.unsw.edu.au>
Currently raid1 uses PAGE_SIZE read/write requests for resync, as it doesn't
know how to honour per-device restrictions. This patch uses to bio_add_page
to honour those restrictions and ups the limit on request size to 64K. This
has a measurable impact on rebuild speed (25M/s -> 60M/s)
Andrew Morton [Wed, 18 Feb 2004 12:51:40 +0000 (04:51 -0800)]
[PATCH] md: Avoid unnecessary bio allocation during raid1 resync
From: NeilBrown <neilb@cse.unsw.edu.au>
For each resync request, we allocate a "r1_bio" which has a bio "master_bio"
attached that goes largely unused. We also allocate a read_bio which is
used. This patch removes the read_bio and just uses the master_bio instead.
This fixes a bug wherein bi_bdev of the master_bio wasn't being set, but was
being used.
We also introduce a new "sectors" field into the r1_bio as we can no-longer
rely in master_bio->bi_sectors.
Andrew Morton [Wed, 18 Feb 2004 12:51:31 +0000 (04:51 -0800)]
[PATCH] md: Remove some un-needed fields from r1bio_s
From: NeilBrown <neilb@cse.unsw.edu.au>
next_r1 is never used, so it can just go.
read_bio isn't needed as we can easily use one of the pointers in the
write_bios array - write_bios[->read_disk]. So rename "write_bios" to "bios"
and store the pointer to the read bio in there.
Andrew Morton [Wed, 18 Feb 2004 12:51:22 +0000 (04:51 -0800)]
[PATCH] md: Discard the cmd field from r1_bio structure
From: NeilBrown <neilb@cse.unsw.edu.au>
The only time it is really needed is to differentiate a retry-on-fail from a
write-after-read-for-resync request to raid1d. So we use a bit in 'state'
for that.
Andrew Morton [Wed, 18 Feb 2004 12:51:13 +0000 (04:51 -0800)]
[PATCH] md: Split read and write end_request handlers
From: NeilBrown <neilb@cse.unsw.edu.au>
Instead of having a single end_request handler that must determine whether it
was a read or a write request, we have two separate handlers, which makes
each of them easier to follow.