| 1 |
-----BEGIN PGP SIGNED MESSAGE----- |
| 2 |
Hash: SHA1 |
| 3 |
|
| 4 |
Hi, |
| 5 |
|
| 6 |
I had problems with CFS too. Similar to the one's you describe. Under |
| 7 |
heavy disk/cpu-load (like emerge and updatedb in parallel) the system |
| 8 |
was unresponsive. I changed the scheduler back to anticipatory and the |
| 9 |
problem was gone. |
| 10 |
Interestingly i also have a thinkpad, X30 in my case. |
| 11 |
|
| 12 |
best |
| 13 |
Thomas |
| 14 |
|
| 15 |
fire-eyes wrote: |
| 16 |
| Andrey Falko wrote: |
| 17 |
|> On Thu, Mar 6, 2008 at 10:38 PM, fire-eyes <sgtphou@×××××××××.org> wrote: |
| 18 |
|>> Hello, |
| 19 |
|>> |
| 20 |
|>> I'm running gentoo on an IBM Thinkpad T43. In the past, disk access was |
| 21 |
|>> fine under load (CPU or disk). These days, the disk becomes painfully |
| 22 |
|>> slow under moderate to high CPU usage (such as compiling) or copying |
| 23 |
|>> more than a few MB. GUI's become almost totally unresponsive and at |
| 24 |
|>> times I have to down the system hard. |
| 25 |
|>> |
| 26 |
|>> So it seems it's some sort of a change in kernels compared to the past. |
| 27 |
|>> I have always run a vanilla kernel, manually configured and installed. |
| 28 |
|>> Right now I am running 2.6.24.3. |
| 29 |
|>> |
| 30 |
|>> The system uses an SATA disk drive. |
| 31 |
|>> |
| 32 |
|>> Here is the boot line in grub.conf: |
| 33 |
|>> kernel /boot/vmlinuz-stable root=/dev/sda4 rw hdc=noprobe |
| 34 |
|>> acpi_sleep=s3_bios panic=5 elevator=cfq nmi_watchdog=0 |
| 35 |
|>> |
| 36 |
|>> /boot/vmlinuz-stable being a symlink to the kernel I consider "stable" |
| 37 |
|>> within /boot/. I also have vmlinuz-last called by another grub entry if |
| 38 |
|>> I need it. |
| 39 |
|>> |
| 40 |
|>> Here is my kernel config: http://fire-eyes.org/t/config-2.6.24.3.txt |
| 41 |
|>> (may disappear in the future) |
| 42 |
|>> |
| 43 |
|>> I am looking for feedback into what may be causing this mess. It makes |
| 44 |
|>> for a very frustrating time using this laptop. |
| 45 |
|>> |
| 46 |
|> What is the version of the kernel where you did not have issues? |
| 47 |
|> 2.6.24 and 23 have a new CPU scheduler (CFS), which "should" work |
| 48 |
|> better than the old one. It is possible that the new scheduler does |
| 49 |
|> not suit your needs. |
| 50 |
| |
| 51 |
| Thanks for the reply. |
| 52 |
| |
| 53 |
| I do not recall, other than it was four or more months ago. Do you |
| 54 |
| happen to know what version of the kernel that scheduler showed up in? |
| 55 |
| Also, is that scheduler not irrelevant here as I was passing elevator=cfq? |
| 56 |
| |
| 57 |
| By the way, I did a little experimentation. I changed my scheduler to |
| 58 |
| deadline, and set preemption to desktop. Before the scheduler was cfq, |
| 59 |
| and the preemption to low-latency desktop. |
| 60 |
| |
| 61 |
| Things already feel snappier gui-wise, but I have yet to push the |
| 62 |
| disk/cpu to see what will happen. I believe it is at least the start of |
| 63 |
| improvements, however. |
| 64 |
|
| 65 |
-----BEGIN PGP SIGNATURE----- |
| 66 |
Version: GnuPG v2.0.7 (GNU/Linux) |
| 67 |
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org |
| 68 |
|
| 69 |
iD8DBQFH0P8nrpEWPKIUt7MRAqXHAJ96MDsVC0xAwm2f/5uSWbQLxZLLsgCdELyO |
| 70 |
WnJkehZ+0MSsujfd1vaSrPY= |
| 71 |
=WA+M |
| 72 |
-----END PGP SIGNATURE----- |
| 73 |
-- |
| 74 |
gentoo-user@l.g.o mailing list |
Archives