Jochen Suckfuell wrote:
: However, since you don't have SCSI disks, this will most likely not fix
: your problem. But before I was able to track down this missing lock
: issue, I could fix the problem by making the ios_in_flight variable use
: atomic operations. Let me guess, you have an SMP machine, where a race
: condition is likely to occur?
Nope. Single-CPU Athlon 850, Via KT133 (not KT133A), ASUS
board with on-board VIA IDE and on-board promise controller, 1.1GB RAM
(CONFIG_HIGHMEM), 3c985B NIC.
: Below is my patch to use atomic operations (against 2.4.19-pre10) that
: fixed the problem for me on an SMP machine with SCSI disks. (I don't
: have a non-SCSI SMP machine to check.)
:
: In the first chunk, I also added a check for negative ios_in_flight
: values to the function generating the /proc/partitions output, and
: eventually set them to zero. You should remove this part to see if
: values are still becoming negative.
:
: If this patch works, there must be another place where accounting is
: done without proper spinlock handling. The atomic-patch should not be
: the final solution for this!
I'll look at this, but I am about to leave for a holiday,
so don't expect feedback in next two weeks or so. Also I don't want to
experiment with this server more than necessary, and this is not a critical
bug.
: rio/wio are the total number of read/write requests processed, while
: rmerge/wmerge is the number of successful merges of requests.
What does it mean "merges" and how is it related to the
mult-count of the HDD? I've tried to run "dd if=/dev/hdc bs=4k of=/dev/null"
on a drive with mult-count 16 and read-ahead 8, and the rsect count
was approximately 126.5 times bigger than rio, while the rsect/rmerge
was about 62. I've changed the "bs=4k" to "bs=64k" in the "dd" command,
but the rsect/rio and rsect/rmerge ratio was about the same.
So how does the read() syscall from dd(1) maps to rio requests
(and rmerge requests)? Is it related to the drive multcount at all?
-Yenya
--
| Jan "Yenya" Kasprzak <kas at {fi.muni.cz - work | yenya.net - private}> |
| GPG: ID 1024/D3498839 Fingerprint 0D99A7FB206605D7 8B35FCDE05B18A5E |
| http://www.fi.muni.cz/~kas/ Czech Linux Homepage: http://www.linux.cz/ |
|----------- If you want the holes in your knowledge showing up -----------|
|----------- try teaching someone. -- Alan Cox -----------|
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Jul 30 2002 - 14:00:24 EST