These are the missing patches from me:
SA_INTERRUPT-shared-1: avoids SA_INTERRUPT to break SA_SHIRQ.
SMP-boot-race-2.2.10-B: avoid the first CPU to launch the first kernel
thread (so calling the scheduler code) before
the other CPUs had a chance to enter sys_idle()
to set cpu_curr() != NULL.
bh-SMP-races-2.2.10-A: fixes the bh_mask_count/bh_mask smp races (ala
sparc).
buffer-races-2.2.10-A: fixes genuine race in set_blocksize and
invalidate_buffers and make sure the
invalidate buffers are placed in the freelist.
endbase-2.2.10: safely heuristic to read the EDBA register from
the bios (old hardware will run fine as default).
If you don't have a bios you can override via
kernel param.
flush-tlb-i386-1: invlpg if not present must be replaced with
a %%cr3 noop write and not with a flush of
the current task.
hashed-buffers-2.2.10: really a minor issue, the right place for
updating the hashed-buffers stats is in the
path that really remove the buffer from the
hashtable.
no-swapout-2.2.10-B: allow 2.2.12 to run on real servers without
swapin/swapouts all the time (for example during
heavy filesystem activity) that make 2.2.x
unuseable otherwise (people was using 2.0.x before
this my fix).
ntp-SMP-races-1: fix for settimeofday/stime/adjtimex
SMP races. I don't claim it to fix the
fix for RTC screwedup with ntpd on SMP but it's
sure good to have such bug fixed too.
oom-2.2.12-I: Fix a bad deadlock condition in the swapin
code that may soft-deadlock the machine during
swapout activities if the swapin is triggered
inside the kernel and the swapin can't allocate
the page. It's pretty easy to trigger this race
simply generating some syslog message (for example
logging in via network) while the machine is going
to sigkill a task in oom(). It also fixes the
SIGBUS accessing a shared mapping beyond the
end of the file on Alpha. Also avoids iopl()
application to be SIGKILLED and send a
graceful SIGTERM that avoid screwing up the
console by going out of memory.
This breaks all architectures but it's strictly
necessary. I fixed only Alpha and i386 so far
as I have no other hardware.
probe-irq-2.3.14-pre2-1: avoid adding workarounds in the code by probing
two times for irqs. The problem is that an irq
may be pending just a bit before probing irq
and so such pending irq was mistaken for
a spurious irq so causing the probe to fail.
shrink_all_cache-2.2.10-A: set to zero the minimum limit of cache and
buffers. This goes with the no-swapout patch
for big memory servers.
smp-reschedule-boot-2.3.13-1: fix a race in the SMP boot that was
causing the master CPU to reschedule the
future dummy idle task of a secondary
cpu instead of continuing the boot stage
(so i386 8 (or more) way SMP will
boot fine without hanging in wait-for-callout)
wakeup_bdflush-2.2.10-A: avoid a deadlock in wakeup_bdflush.
run_task_queue() will sleep if we are flushing
loop buffers to disk.
You can find all the patches here:
ftp://ftp.suse.com/pub/people/andrea/kernel-patches/my-2.2.12/
(if you want them by email just drop me a note)
Andrea
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/