Hi,
I think that the problem is caused by ehci driver instead of ARM SMP
scheduler, and I
have verified that patch below can fix the problem. And the patch has
been posted
on usb/arm/omap mail list for discussion.
But I am wondering that why no such problem after passing 'nosmp' to
kernel.
** Also affects: linux-ti-omap4 (Ubuntu Natty)
Importance: Undecided
Status: New
** Also affects: linux-ti-omap4 (Ubuntu Maverick)
Importance: Undecided
Status: New
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
** Description changed:
+ Original Bug name: panda: USB disk IO slow
+
My Panda's USB seems to be significantly slower than a Beagle C4.
hdparm shows buffered reads as ~12MB/s on the Panda, and about ~20-25MB/s on
a Beagle C4 from the same
external Lacie USB disk.
Kernel is
I ran the tests again, with the noop, cfq and deadline I/O schedulers,
and the results differed in interesting ways. See attached logs.
** Attachment added: noop test run 1
https://bugs.launchpad.net/linux-linaro/+bug/709245/+attachment/2256614/+files/speedtest-noop.log
--
You received this
** Attachment added: cfq test run 1
https://bugs.launchpad.net/linux-linaro/+bug/709245/+attachment/2256615/+files/speedtest-cfq.log
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
** Attachment added: deadline test run 1
https://bugs.launchpad.net/linux-linaro/+bug/709245/+attachment/2256616/+files/speedtest-deadline.log
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
** Attachment added: noop test run 2
https://bugs.launchpad.net/linux-linaro/+bug/709245/+attachment/2256617/+files/speedtest-noop%20%282%29.log
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
** Attachment added: cfq test run 2
https://bugs.launchpad.net/linux-linaro/+bug/709245/+attachment/2256618/+files/speedtest-cfq%20%282%29.log
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
** Attachment added: deadline test run 2
https://bugs.launchpad.net/linux-linaro/+bug/709245/+attachment/2256619/+files/speedtest-deadline%20%282%29.log
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
Maybe the 1ms threshold thing is just to do with a threshold heuristic
where it stops using the scheduler and sleeping and instead spins
udelay() style, hence the 100% CPU. That's not to write off its effect
on the bug just a possible explanation of why it cares about 1ms or
less.
Reading these
Can someone check does it happens on other multicore A9 board? Origen or
Versatile Express (Snowball lacks usb host port).
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
panda: USB
Let me quote from #pandaboard:
prpplague TheSeven: the direction the debugging is going is that this is a
thread issue, i suspect that the .0099 is just under the thread timing
prpplague TheSeven: linaro folks and fedora folks are working on debugging
this
prpplague TheSeven: it doesn't appear
I just tried the 3.0 kernel, and it behaves in exactly the same way.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
panda: USB disk IO slow
To manage notifications about this bug go
I tested some more today (Oneiric 3.0 kernel) and the external USB drive
on a Panda.
With nosmp passed on the boot command line, the throughput is good with
or without the ping and about the same as using both cores and pinging
it.
Booting with both cores and putting cpu1 offline does not seem
** Tags added: iso-testing
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
panda: USB disk IO slow
To manage notifications about this bug go to:
I see the same behaviour without ethernet plugged in, and pinging the
panda via wlan0. Is that chip too on the same USB controller?
5x speed increase on write when pinging with -i.0001
jani@panda:~$ time dd if=/dev/zero of=alma bs=4M count=1024
1024+0 records in
1024+0 records out
4294967296
Hi Jani -
No, Panda WLAN is on a 4-bit SDIO interface. So it's completely
different from ping on Ethernet, I don't know what that means that it
still changes USB throughput.
Also when I reproduced this via pinging ethernet, I only saw a 50%
increase in throughput; maybe it's to do with the
Guess what, this issue indeed seems to be completely unrelated to USB.
Something similar is going on for SD card access (see Bug #787246), and
accessing LAN/WLAN has only an indirect influence.
Look at this: http://paste.ubuntu.com/656410/
Seems like the CPU entering sleep states or something
On Mon, 1 Aug 2011, Michael Sparmann wrote:
Guess what, this issue indeed seems to be completely unrelated to USB.
Something similar is going on for SD card access (see Bug #787246), and
accessing LAN/WLAN has only an indirect influence.
Look at this: http://paste.ubuntu.com/656410/
Seems
No noticable improvement: http://paste.ubuntu.com/656498/
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
panda: USB disk IO slow
To manage notifications about this bug go to:
On Mon, 1 Aug 2011, Michael Sparmann wrote:
No noticable improvement: http://paste.ubuntu.com/656498/
The nohlt argument prevents the idle power saving code from being called
at all. So if that doesn't help the USB case either than it is unlikely
that anything related to sleep modes.
--
Further testing points towards thread switching behavior being related:
Nothing running in parallel: 10MB/s at big block sizes
perl -e while (1) { }: 5MB/s
perl -e use Time::HiRes qw( usleep ); while (1) { usleep(1) }: 20-25MB/s
perl -e use Time::HiRes qw( nanosleep ); while (1) { nanosleep(1) }:
I have added nohlt and seems no any changes wrt. usb storage performance.
But the following test is interesting, just run a 'busy' in background and
increas usb
storage performance triple or 4 times, and killing the 'busy' task will make
usb
storage performance poor like before.
see my test
Another interesting thing, enabling usbmon will increase performance a lot.
root@tom-panda:~# dd if=/dev/sda iflag=direct of=/dev/null bs=128K count=32
32+0 records in
32+0 records out
4194304 bytes (4.2 MB) copied, 3.4598 s, 1.2 MB/s
root@tom-panda:~# cat /sys/kernel/debug/usb/usbmon/1u
When running just dd at 1MB block size, I get ~5.5MB/s.
When running with sudo ping -q -i.0100 localhost in the background, ping causes
1% CPU load, and I get ~10MB/s at 1MB block size and these ping statistics:
6228 packets transmitted, 6227 received, 0% packet loss, time 97329ms
rtt
** Tags added: armel kernel-bug
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
panda: USB disk IO slow
To manage notifications about this bug go to:
** Also affects: linux-ti-omap4 (Ubuntu Oneiric)
Importance: High
Status: Confirmed
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
panda: USB disk IO slow
To manage
** Changed in: linaro-ubuntu
Milestone: None = 11.07
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/709245
Title:
panda: USB disk IO slow
To manage notifications about this bug go to:
Appears to still be a problem with 3.0.0-1:
...
Threaded I/O Tester:
pts/tiobench-1.1.0 [Test: Write - Size Per Thread: 32MB - Thread Count: 4]
Test 1 of 4
Estimated Time Remaining: 4 Hours, 57 Minutes
Estimated Test Run-Time: 28 Minutes
Expected Trial Run Count: 4
I was able to reproduce this in Maverick as well.
** Attachment added: bonnie-maverick.html
https://bugs.launchpad.net/ubuntu/+source/linux-ti-omap4/+bug/709245/+attachment/2221456/+files/bonnie-maverick.html
--
You received this bug notification because you are a member of Ubuntu
Bugs,
Would be good to check if you're able to reproduce this issue with the
3.0 based kernel.
** Also affects: linaro-ubuntu
Importance: Undecided
Status: New
** Changed in: linaro-ubuntu
Status: New = Confirmed
** Changed in: linaro-ubuntu
Importance: Undecided = High
--
You
** Also affects: linux-ti-omap4 (Ubuntu)
Importance: Undecided
Status: New
** Changed in: linux-ti-omap4 (Ubuntu)
Importance: Undecided = High
** Changed in: linux-ti-omap4 (Ubuntu)
Status: New = Confirmed
--
You received this bug notification because you are a member of
I have been able to reproduce this bug using bonnie++. The process I
used was to install Oneiric using the netboot installer to an external
USB SATA drive and boot with that as my root filesystem (SD only has
bootloaders kernel).
Steps to reproduce on a running system:
On panda (no ping):
sudo
33 matches
Mail list logo