Problem analysis and fix provided by Andries Brouwer.
Cleanup of nfs_lookup() by me...

into fys.uio.no:/home/linux/bitkeeper/nfsclient-2.5

ESTALE errors.

resent since you don't know whether the server is replying to the
first or the second transmission. However we're currently setting
the cutoff point to be the timeout of the first transmission.

This patch moves the cutoff point to the actual start of the
retransmission.

in line with what is done for TCP.

Signedness corrections when updating RTT.

Be conservative when calculating RTO. Round up the residues.

among other things, for the "soft" mount option.

that call_encode() may cause rq_rcv_buf to be reset despite the fact
that a reply might be delivered at any moment by a softirq.
   
This typically results in 'NFS: server cheating in read reply'
error messages.   

Solve by adding rq_private_buf, which is updated atomically from
rq_rcv_buf.

   - Ensure that we reset req->rq_received before resending if the
     server has sent us a garbage reply.

   - Whenever we grab the xprt_lock_write "semaphore" ensure that we
     reset req->rq_bytes_sent.

   - When resending a TCP request, do not interrupt in the
     middle of transmission even if we do get a reply from the server.

   - Protect the call to rpc_calc_rto() against modification while
     we are calculating

Fix APIC ID lookup.  In the bios_cpu_apicid[] case, it would walk off
the end of bios_cpu_apicid[] and attempt to send APIC INIT messages to
garbage without this patch, and in the NUMA-Q case, it would attempt to
send NMI wakeups to destinations in the broadcast cluster (which is
harmless, but very poor form) without this patch.

The MSRs might change between CPU models, error out if we haven't met
the CPU before. From Venkatesh Pallipadi

Add a stat counting the (relatively common) case where a PC value is logged
but there is no (longer) any executable mapping covering that address.

Tell user-space how big kernel pointers are, as preferable to sniffing /proc/kcore.
Improve the oprofilefs_ulong_to_user() prototype.

Andi Kleen pointed out the MSRs array was a massive bloat source. Reduce
it somewhat by only allocating the amount actually needed for the CPU type.

Untested on Pentium IV - I don't have a machine.

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

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

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

remove spurious logging of message on "create if file does not exist" case (without O_EXCL) when file exists

When addrconf_ifdown() calls ipv6_mc_destroy_dev(), it has NULL'd
out dev->ip6_ptr, which means all in6_dev_get() calls will fail.
So pass an explicit idev into ipv6_dev_mc_dec() in this case so
that we don't leak the all-nodes multicast address reference to the
idev.

From Dominik Brodowski.

- Can simplify, since the policy is already known, and we only need to kick
  the CPU.

- Make sure we set the right method in the sysdev driver.

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

We don't need to protect against multiple suspend calls, since subsequent
ones will happen on an empty (or at worst - short) list.

into osdl.org:/home/mochel/src/kernel/linux-2.5-power

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

Calls were moved to the PM core, so they must be compiled in to use them.

- Didn't realize that mount_root_failed_msg() had been removed, so it was 
  accidentally merged back in. It's gone again now, as well as the entire
  (unused) dmi_dump_system() function. 

Dan hasn't been maintaining cramfs for some time.

Mark it orphan until somebody else takes it up.

- Revert names of functions back to device_{suspend,resume} since at least 
  APM are still using them.

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

This removes the pointless cramfs message when booting with cramfs
compiled in but not as the root filesystem.

into osdl.org:/home/mochel/src/kernel/linux-2.5-power

into kernel.bkbits.net:/home/davem/net-2.5

Cleanup the timer code.  It works as is, but:
- use timer_ macro's when working with jiffies.
- cleanup the #define's
- make timer run sooner (ie when next entry will expire)
  rather than waiting till the next 10 minute quanta.
- print TTL value in seconds not jiffies
- change formatting to linux style in code the area affected.