Hello :) Just wanted to see if any of you are experiencing the same problem - likely this is a bug?
I am using a USB 2.0 disk (a Seagate 120 GB disc) - ST3120022A, also
reported as:
Host: scsi1 Channel: 00 Id: 00 Lun: 00
Vendor: ST312002 Model: 2A Rev: 0 0
Type: Direct-Access ANSI SCSI revision: 02
I use the uhci module (but I tried both uhci modules, the effect is the
same with both of them).
I have great problems doing long transfers to this device. After a time
the transfer becomes really slow - or stops for some time, only to
continue a bit later. But if I try that a few times, or with large files
( more than 400 MB, or something like that several times) in the end my
system hangs.
I notive a _VERY_ high memory consumption during a transfer.
At the end of the transfer of a 400 MB file, free reports this:
total used free shared buffers cached
Mem: 774400 770052 4348 0 8120 659056
-/+ buffers/cache: 102876 671524
Swap: 524624 48 524576
I see my hard disk working for quite some time, even after the transfer is
apparently over (cp returns control to the terminal). dmesg also still
shows activity. (see attached file: dmesg.out.new.2) This goes on for
several more minutes (Definitely more than 5).
After the drive finally signals no more activity, the memory situation is
a bit better:
total used free shared buffers cached
Mem: 774400 767224 7176 0 9432 656316
-/+ buffers/cache: 101476 672924
Swap: 524624 48 524576
...this leads to problems with the next data transfers (I assume, since my
PC hangs even faster with the 2nd copy to the usb-harddisk). Only
unmounting the disk helps.
I also noticed this problem when creating filesystems on the disk.
Creating filesystems bigger than 20 or 30 GB is a real problem - it takes
forever and, like the transfer of a file, has a chance of hanging my pc.
Usually I'd say this is not a problem, since this is only cached memory
and should be available for applications, but this doesn't really seem to
be the case here. Maybe, however this is only a symptom and not the
problem? :)
Anyone else has seen this problem? :)
Is there a workaround, need more information, or can someone tell me
what's happening there?
Oh - my kernelversion is 2.4.22pre5, and I am at this moment using the
alternate uhci driver.
The dmesg.out file shows the initial state, after booting as the drive is
logged in.
dmesg.out.new.2 shows the communication, even after the copy actions seems
to have finished...
--
Today is Prickle-Prickle, the 48th day of Confusion in the YOLD 3169
dmesg.out
Description: Binary data
dmesg.out.new.2
Description: Binary data
pgp00000.pgp
Description: PGP signature
