Re: v3.5 nfsd4 regression; utime sometimes takes 40+ seconds toreturn

[Jamie Heilman]

Jamie Heilman wrote:
> I'll try to get full rcpdebug traces on client and server as the delay
> is occuring in the hopes that helps pin things down, and post them
> separately.

OK, here are the logs from client and server, where a run of my test program
under strace -T resulted in:

utime("utime-test.c", [2012/08/15-13:35:20, 2012/08/15-13:22:04]) = 0 <0.150815>
open("utime-test.c", O_RDONLY)          = 3 <0.242635>
close(3)                                = 0 <0.147768>
stat("utime-test.c", {st_mode=S_IFREG|0644, st_size=696, ...}) = 0 <0.002772>
utime("utime-test.c", [2012/08/15-13:35:20, 2012/08/15-13:22:04]) = 0 <71.878058>

The client is a 64-bit v3.4.8 kernel, the server is 32-bit, v3.5.1 +
the two sunprc patches that will be in v3.5.2.

(The client's system clock is a touch faster than the server's, but
these logs start at the same instant.)


-- 
Jamie Heilman                     http://audible.transient.net/~jamie/

Attachment: server.gz
Description: Binary data

Attachment: client.gz
Description: Binary data