into home.osdl.org:/home/torvalds/v2.5/akpm

From: James Bottomley <James.Bottomley@SteelEye.com>

The compiler justly complains in this:

	        unsigned int regnr = *(loff_t *) v, i;

	        if (regnr == 0) {
	                seq_puts(p, "     ");
	#ifdef CONFIG_SMP
	                for (i = 0; i < NR_CPUS; i++)
	#endif
	                        seq_printf(p, "      CPU%02d ", i);

That i is uninitialised if CONFIG_SMP is not set.

From: Ram Pai <linuxram@us.ibm.com>

I have enclosed a patch that fixes a bunch of performance bugs in the
readahead code.

Below is a brief summary of the problems noticed and the proposed fixes
with some results:
   
 Problem 1:  Readahead code closes the readahead window and goes into
 slowread path, if a file is accessed the first time at an offset
 notequal to zero.  In the case of databases(especially in db2), a file
 may not be accessed at offset 0 the first time though the i/o's are
 sequential.

 Fix to Problem 1:
          min = get_min_readahead(ra);
         orig_next_size = ra-next_size;

 -       if (ra-next_size == 0 && offset == 0) {
 +       if (ra-next_size == 0) {

------------------------------------------------------------------------

 ------------------------------------------------------------------------
 Problem 2: After fixing Problem, the readahead window still does not
 open up the first time, if all the pages requested are already in the
 page cache. This time the window closes because of pagecache hits
 instead of misses. To fix this we put in these changes.

 -               check_ra_success(ra, ra-size, actual, orig_next_size);
 +               if(!first_access) {
 +                       check_ra_success(ra, ra-size, actual, orig_next_size);
 +               }

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------
 Problem 3: In the case of large random reads, the readahead window is
 read in, the moment there is a hit in the active window. And it turns
 out that in most of the cases the readahead window gets scrapped,
 because the next large random read does not even touch any of the pages
 in that readahead window. We fixed this by introducing lazy readahead.
 Basically we wait till the last page in the active window gets a hit.
 And once the last page is hit, the readahead window is then read in.
 This fix gave a tremendous boost in the performance.
 To fix this the changes we put in were:

                 /*
                  * This read request is within the current window.  It is time
                  * to submit I/O for the ahead window while the application is
                  * crunching through the current window.
                  */
 -               if (ra-ahead_start == 0) {
 +               if (ra-ahead_start == 0 && offset == (ra-start + ra-size -1)) {

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------
 Problem 4:
       If the request page does not fall in the active window and is not
the  first page of the read ahead window, we scrap both the active
window and the readahaed window and read in the active window. But it
turns out that we read in a lot of pages in the active window based on
the size of the 'projected readahead window size' (the next_size
variable). And we end up using part of the active window and waste the
remaining. We put  in a fix where we read in just as many pages in the
active window based on the number of pages used in the recent past.
 Again this gave us another big boost in performance and ended up
beating the performance of aio patch on a DSS workload.

 The fix to this is:
                  * ahead window and get some I/O underway for the new
                  * current window.
                  */
 +               if (!first_access && preoffset = ra-start &&
 +                               preoffset < (ra-start + ra-size)) {
 +                       ra-size = preoffset - ra-start + 2;
 +               } else {
 +                       ra-size = ra-next_size;

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------
 Problem 5:
 With all the above fixes there is very low chance that the readahead
 window shall close.  But however if it does, we found that the slow
 read  path is really slow.  Any loss of sequentiality in the slow read
 path is penalized heavily by closing the window back to zero. So we
 fixed this  by decreasing the window size by one anytime we loose
 sequentiality and  increasing in by 1 if we didn't.

                 if (offset != ra-prev_page + 1) {
 -                       ra-size = 0;            /* Not sequential */
 +                       ra-size = ra-size?ra-size-1:0; /*Notsequential */

 ------------------------------------------------------------------------

With the above set of fixes we got about 28% improvement in DSS workload
which is about 5% more than what we got with the suparna's aio patch.

This patch compared equivalent to suparna's aio patch with aio-stress
run.

It fared better than aio patch for large random io.

We are yet to run a bunch of other benchmarks to evaluate this patch.
We would like to get your inputs on this patch and any suggestions you
may have to improve it. I have enclosed a patch with all these changes
along with some changes to the comments that reflect the new behaviour.
NOTE: the above patch reverts suparna's aio patch.

From: Hans Ulrich Niedermann <linux-kernel@n-dimensional.de>

I've noted that 2.6.0 contains broken references to documentation.

I got sufficiently annoyed chasing doc files in the wrong place
that I wrote a script to check the references to documentation
files.

Some documentation files have moved (e.g.  Documentation/modules.txt to
Documentation/kbuild/modules.txt).  I adapted the references with a script.

From: Nikita Danilov <Nikita@Namesys.COM>

This patches reiserfs_rename.  It adds ctime update of renamed object.  It
also fixes calculation of maximal possible transaction size during rename.

Thanks to Alex Adriaanse <alex_a@caltech.edu> for finding this.

From: Nikita Danilov <Nikita@Namesys.COM>

Add "commit" reiserfs mount option to override maximal transaction age. 
Usage:

mount -treiserfs -ocommit=<time-in-seconds> /device /mountpoint

Submitted by Hugang <hugang@soulinfo.com>.

From: Nikita Danilov <Nikita@Namesys.COM>

Patch to teach fs/reiserfs/super.c:reiserfs_fill_super() to respect @silent
parameter and to not issue any output if @silent is set.

Also remove some trailing white spaces, while we are here.

From: Muli Ben-Yehuda <mulix@mulix.org>

It's messy, and needs describing.

From: BlaisorBlade <blaisorblade_spam@yahoo.it>

This fixes a "typo" for Kconfig-language docs.

wli points out that shrink_slab inverts the sense of shrinker->seeks: those
caches which require more seeks to reestablish an object are shrunk harder.
That's wrong - they should be shrunk less.

So fix that up, but scaling the result so that the patch is actually a no-op
at this time, because all caches use DEFAULT_SEEKS (2).

log_buf_len_setup() is called on the start_kernel->parse_args() path.  It
must not enable interrupts.

From: Harald Welte <laforge@gnumonks.org>

[disclaimer:  This was posted on the linuxppc list before, BenH asked me=20
 to re-post it to lkml]

The prism54 (http://prism54.org) driver for my cardbus adapter works
with 2.4.x, but not 2.6.x on a Titanium G4 Powerbook IV.

On 2.6.x the error message was
PCI:0001:02:00.0 Resource 0 [00000000-00001fff] is unassigned

After investigating differences in the PCI code of 2.4.x and 2.6.x, i
noticed that 2.4.x/arc/ppc/kernel/pci.c:pcibios_update_resource()
contained a couple of lines that unset the IORESOURCE_UNSET bitflag.

In 2.6.x, this is handled by the generic PCI core in
drivers/pci/setup-res.c:pci_update_resource() code.  However, the code
is missing the 'res->flags &=3D ~IORESOURCE_UNSET' part.

The below fix re-adds that section from 2.4.x.=20

I'm not sure wether this belongs into the arch-independent PCI api.
Anyway, on PPC it seems to be needed for certain cardbus devices.

Any comments welcome.

qlogic's 16k maximum I/O size bites again.  Neil says:

 The problems we fixed were all to do with normal IO.  This one is resync
 IO.

 The problem here is that raid1 always does resync in RESYNC_BLOCK_SIZE
 (=64k) IOs and if the device doesn't cope - tough.

 The simple fix is to #define RESYNC_BLOCK_SIZE to PAGE_SIZE in md/raid1.c

 The better fix is to rewrite the raid1 resync code to use bio_add_page.

 This means we have to build the read request and the write requests at the
 same time, and then when a bio_add_page fails, we back-out the last page
 from the bios that have one too many, and then do the read followed by the
 writes.

 I have some code the nearly does this, but I haven't got it actually
 working yet and I am on leave until mid January.

 I would recommend doing

 fs/drivers/md/raid1.c:
 -#define RESYNC_BLOCK_SIZE (64*1024)
 +#define RESYNC_BLOCK_SIZE PAGE_SIZE

 for now.

From: Jon Burgess <mplayer@jburgess.uklinux.net>

The iso9660 filesystem code checks that the "sbsector" option is
positioned within the first 660Mb of the disk.

Today the iso9660 filesystem is used on DVD's which are much bigger than
660Mb and this check prevents the sbsector option being used to specify 
the location of the superblock of multisession DVD's.

With this check removed I can mount the second session on a DVD-R by 
specifying the sbsector, even though the  firmware on that drive returns
bogus data for the TOC.

If an invalid large sector number is entered then a "request beyond end 
of device" error is reported elsewhere in the block code, but appears to
do no damage.

From: Alan Cox <alan@redhat.com>

Change by Clear Zhang of ALI

From: Bernardo Innocenti <bernie@develer.com>

It needs init.h

From: Steven Cole <elenstev@mesatop.com>

From: "Seth, Rohit" <rohit.seth@intel.com>

We recently covered a bug in mm/mmap.c on IA-64.  While unmapping a address
space, unmap_region calls free_pgtables to possibly free the pages that are
used for page tables.  Currently no distinction is made between freeing a
region that is mapped by normal pages vs the pages that are mapped by
hugepages.  Architecture specific code needs to handle cases where PTEs
corresponding to a region that is mapped by hugepages is properly getting
unmapped.  Attached please find a patch that makes the required changes in
generic part of kernel.  We will need to send a separate IA-64 patch to use
this new semantics.  Currently, so not to disturb the PPC (as that is the
only arch that had ARCH_HAS_HUGEPAGE_ONLY_RANGE defined) we are mapping back
the definition of new function hugetlb_free_pgtables to free_pgtables.

From: Joe Korty <joe.korty@ccur.com>

We need to check whether preemption is needed after reenabling preemption
in kunmap_atomic().

From: Mark Haverkamp <markh@osdl.org>

This fixes a problem similar to the patch I submitted on 11/20

http://marc.theaimsgroup.com/?l=linux-kernel&m=106936439707962&w=2

In this case, though, the result is an:

"Incorrect number of segments after building list" message.

The macro __BVEC_START assumes a bi_idx of zero when the dm code can
submit a bio with a non-zero bi_idx.
The code has been tested on an 8 way / 8gb OSDL STP machine with a 197G
lvm volume running dbt2 test.

From: Juergen Quade <quade@hsnr.de>

I just hit a copy-paste typo in floppy.c.

From: Jim Radford <radford@indigita.com>

I finally took the time to figure our why my parallel port wasn't
working...  here's the patch.

From: Michael Hunold <hunold@convergence.de>

the Linux-on-a-CD system Knoppix has nearly all framebuffer drivers for
2.4.23 compiled in. Additionally, it surpresses most kernel messages by
lowering the kernel log level.

Two framebuffer drivers (clgenfb.c and hgafb.c), however, use KERN_ERR
to say that their particular card has *not* been found which is very
annoying.

Especially the clgenfb.c driver simply says on bootup:
  >  Couldn't find PCI device
which can really confuse newbie users.

I've already send a patch that fixes this for 2.4 -- Marcelo and Geert
Uytterhoeven have already ack'ed it.

The same change should be done for 2.6, too IMHO.

The appended patch replaces two KERN_ERR with KERN_INFO and additionally
makes the cirrusfb.c message more descriptive.

From: "Randy.Dunlap" <rddunlap@osdl.org>

Add Intel SpeedStep zero-page memory usage doc.

From: Yoshinori Sato <ysato@users.sourceforge.jp>

* using generic_ffs
* optimized code

From: IWAMOTO Toshihiro <iwamoto@valinux.co.jp>

I found linux-2.6.0-test11 leaks memory when execve fails.  I've also
checked the bitkeeper tree and the problem seems to be unchanged.

The attached patch is a partial backout of bitkeeper rev.  1.87 of
fs/exec.c.  I guess the original change was a simple mistake.
(free_arg_pages() is a NOP when CONFIG_MMU is defined).

From: "H. Peter Anvin" <hpa@zytor.com>

This patch fixes the "non-terminating inflate" problem that Russell King
complained about on LKML earlier today.

I chose to use "goto" much like zlib does, in order to not require
setjmp/longjmp inside the kernel.  It's a bit ugly, but it also lets each
function chose how it needs to be terminated on error, which is a good
thing.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

- uses just printk() instead of unneeded fat_fs_panic()
- removes the debug printk
- less verbose on error path
- uses correct a error number on error path

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

If path length became zero by fat_striptail_len(), this returns the -ENOENT
as empty path.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

fatfs misc cleanups/fixes.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

This adds the check of count of clusters.  And if it's too big, fat driver
can't handle it.  So doesn't recognize this as fatfs.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

The -1 was documented as "there is no hint", so this patch uses -1 instead
of 0 for FAT32 fsinfo.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

include/linux/msdos_fs.h cleanup

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Fix printk format

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

For atomicity write, adds readv/writev support to FAT.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
      Michal Rokos <m.rokos@sh.cvut.cz>

The problem is: even if vfat_striptail_len() counts len of name without
trailing dots and sets len to the correct value, utf8_mbstowcs() doesn't
care about len and takes whole name.  So dirs and files with dots can be
created on vfat fs.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
      Yokota Hiroshi <yokota@netlab.is.tsukuba.ac.jp>

This patch is required for my 640MB Optical disk.  Because MS windows 95/ME
based FAT filesystem disk formatter generetes wrong super bloacks.

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

- remove unused "rows" and "cols"
- change the 2 variable to static

From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Suppress a buffer_error() warning which occurs when a page which previously
had an I/O error gets its buffers stripped.

From: David Mosberger <davidm@napali.hpl.hp.com>

This patch makes /proc/PID/maps report the range from FIXADDR_USER_START to
FIXADDR_USER_END as a final pseudo-vma.  This is consistent with the notion
that reading /proc/PID/maps tells you about every page containing data that
the process can in fact access, and with things such as ptrace allowing
access to this memory.  Without this, userland tools that want to look at all
of a process's accessible pages need special-case knowledge about things such
as the vsyscall DSO page.  With this change, existing code that iterates over
the /proc/PID/maps lines will cover those pages like any other.  For example,
this lets gdb's "gcore" command synthesize a core file from a live process
that contains the vsyscall DSO page as a real core dump would, using its
existing generic iterator code and no new special cases.