Bug 12452 (rootdelay_problem)
Summary: | In kernel version 2.6.28, the rootdelay= kernel option does not do anything. | ||
---|---|---|---|
Product: | IO/Storage | Reporter: | Morgan (maclover201) |
Component: | Serial ATA | Assignee: | Jeff Garzik (jgarzik) |
Status: | CLOSED WILL_NOT_FIX | ||
Severity: | high | CC: | akpm, devzero, randy.dunlap |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.28 | Subsystem: | |
Regression: | Yes | Bisected commit-id: | |
Attachments: |
This is the 2.6.28 bootup log from dmesg.
Newer Bootup Log |
Description
Morgan
2009-01-14 17:44:34 UTC
Please attach the full 2.6.28 bootup logs to this report. Created attachment 19813 [details]
This is the 2.6.28 bootup log from dmesg.
(In reply to comment #1) > Please attach the full 2.6.28 bootup logs to this report. > Done. It's attached to this bug. I'm afraid that most of it is missing. It should start with Linux version 2.6.24.3-34.fc8 (mockbuild@xenbuilder4.fedora.phx.redhat.com) (gcc version 4.1.2 20070925 (Red Hat 4.1.2-33)) #1 SMP Wed Mar 12 16:51:49 EDT 2008 Command line: ro root=LABEL=/ rhgb quiet (or similar) You might need to increase the kernel log buffer size to prevent it from overflowing. Add log_buf_len=1M to the kernel boot command line. Thanks. Do you just use dmesg to get that info? Or is it another command? I used dmesg for the previous, incomplete info. If I can do it with dmesg, then I'll add that boot option and get back to you soon. Whoops, accidentially deferred the bug. It's reopened now. ah. The `dmesg' command itself is stupid and truncates the output. Use dmesg -s 1000000 Created attachment 19822 [details]
Newer Bootup Log
This is the new bootup log, although it's not much different. dmesg in busybox doesn't appear to support -s.
> dmesg in busybox doesn't appear to support -s.
argh! We need to see the full logs.
Are they available in /var/log/messages or equivalent?
One alternative might be to disable klogd (rm /sbin/logd or whatever) and
then, after it has booted, do `cat /proc/kmsg > foo'
There is a FULL dmesg log in /var/log. However, it is for the current running kernel. Is there any way to tell a kernel not to write to dmesg for the purpose of extracting the precious text from another kernel? By the way, I am using rootdelay=240 because of another bug that has been triaged by the Ubuntu kernel team. The bug is blocking. I'll post a report on it later, but getting rootdelay working is the first step. Confused. the /var/log file should contain the boot logs for the past several kernels - can't you just extract the one we're after? Or are you saying that something remove the log file on each boot? That would be pretty stupid of them. The other way to fix all this is to get netconsole working: Documentation/networking/netconsole.txt /var/log contains some gzipped logs for previous kernels. They are truncated however. I'll try putting a different version of dmesg that supports -s on a thumb drive, mounting the thumb drive at the initramfs prompt, and running that version. That should do it. so you cannot mount rootfs and want to use rootdelay to fix it ? think we have another issue: [ 3.715643] hda: host max PIO4 wanted PIO255(auto-tune) selected PIO4 [ 3.715925] hda: UDMA/100 mode selected [ 3.716333] Probing IDE interface ide1... [ 34.069528] ide1: Wait for ready failed before probe ! [ 34.070581] CPA self-test: [ 34.070844] 4k 1532 large 253 gb 0 x 1785[c0000000-dfffb000] miss 0 [ 34.084956] 4k 110076 large 41 gb 0 x 110117[c0000000-dfffb000] miss 0 [ 34.095482] 4k 110076 large 41 gb 0 x 110117[c0000000-dfffb000] miss 0 do you see a 30 second realtime delay here ? does it behave different if you disconnect your cd-rom drive (or whatever is also attached to ide1 bus) ? Yeah, I see a delay. I've filed a second bug report here (http://bugzilla.kernel.org/show_bug.cgi?id=12470) because of the delay problem that I have to use rootdelay to work around. The dmesg output pertaining to this issue in 2.6.27.10 is: [ 7.524541] ata1.00: ATA-6: ST380023A, 3.31, max UDMA/100 [ 7.524547] ata1.00: 156301488 sectors, multi 16: LBA [ 7.540518] ata1.00: configured for UDMA/100 [ 12.736015] ata2: link is slow to respond, please be patient (ready=0) [ 17.552015] ata2: SRST failed (errno=-16) [ 22.748015] ata2: link is slow to respond, please be patient (ready=0) [ 27.564015] ata2: SRST failed (errno=-16) [ 32.760017] ata2: link is slow to respond, please be patient (ready=0) [ 62.608015] ata2: SRST failed (errno=-16) [ 67.636015] ata2: SRST failed (errno=-16) [ 67.636023] ata2: reset failed, giving up [ 67.636180] scsi 0:0:0:0: Direct-Access ATA ST380023A 3.31 PQ: 0 ANSI: 5 It's a bit more verbose. Also, disconnecting my CD-ROM drive does nothing. So I did cat /proc/cmdline at the busybox prompt. All that shows up is "term," as opposed to other kernel versions that show all the boot options. Bug closed, I compiled the kernel with the kernel command line overriding boot args. |