From: Kay Sievers <kay@vrfy.org>
To: Joe Perches <joe@perches.com>
Cc: Michael Neuling <mikey@neuling.org>,
linuxppc-dev@ozlabs.org,
Linus Torvalds <torvalds@linux-foundation.org>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Andrew Morton <akpm@linux-foundation.org>,
Steven Rostedt <rostedt@goodmis.org>,
LKML <linux-kernel@vger.kernel.org>, Ingo Molnar <mingo@elte.hu>,
Wu Fengguang <fengguang.wu@intel.com>,
"Paul E. McKenney" <paulmck@us.ibm.com>
Subject: Re: [PATCH v3] printk: Have printk() never buffer its data
Date: Tue, 10 Jul 2012 00:10:17 +0200 [thread overview]
Message-ID: <CAPXgP11Qsj9yGM5nSbRJ_YPFZcr2Km6w4R8wPGU1AytyQChYXQ@mail.gmail.com> (raw)
In-Reply-To: <1341870165.6118.42.camel@joe2Laptop>
On Mon, Jul 9, 2012 at 11:42 PM, Joe Perches <joe@perches.com> wrote:
> On Sun, 2012-07-08 at 19:55 +0200, Kay Sievers wrote:
>
>> At the same time the CPU#2 prints the same warning with a continuation
>> line, but the buffer from CPU#1 can not be flushed to the console, nor
>> can the continuation line printk()s from CPU#2 be merged at this point.
>> The consoles are still locked and busy with replaying the old log
>> messages, so the new continuation data is just stored away in the record
>> buffer as it is coming in.
>> If the console would be registered a bit earlier, or the warning would
>> happen a bit later, we would probably not see any of this.
>>
>> I can fake something like this just by holding the console semaphore
>> over a longer time and printing continuation lines with different CPUs
>> in a row.
>>
>> The patch below seems to work for me. It is also here:
>> http://git.kernel.org/?p=linux/kernel/git/kay/patches.git;a=blob;f=kmsg-merge-cont.patch;hb=HEAD
>>
>> It only applies cleanly on top of this patch:
>> http://git.kernel.org/?p=linux/kernel/git/kay/patches.git;a=blob;f=kmsg-syslog-1-byte-read.patch;hb=HEAD
>>
>
> Hi Kay.
>
> I just ran a test with what's in Greg's driver-core -for-linus branch.
>
> One of the differences in dmesg is timestamping of consecutive
> pr_<level>("foo...)
> followed directly by
> pr_cont("bar...")
>
> For instance: (dmesg is 3.4, dmesg.0 is 3.5-rc6+)
>
> # grep MAP /var/log/dm* -A1
> dmesg:[ 0.781687] ata_piix 0000:00:1f.2: MAP [ P0 P2 P1 P3 ]
> dmesg-[ 0.781707] ata2: port disabled--ignoring
> --
> dmesg.0:[ 0.948881] ata_piix 0000:00:1f.2: MAP [
> dmesg.0-[ 0.948883] P0 P2 P1 P3 ]
>
> These messages originate starting at
> drivers/ata/ata_piix.c:1354
>
> All the continuations are emitted with pr_cont.
>
> I think this output should still be coalesced without
> timestamp deltas. Perhaps the timestamping code can
> still be reworked to avoid too small a delta producing
> a new timestamp and another dmesg line.
Hmm, I don't see that.
If I do:
pr_info("[");
for (i = 0; i < 4; i++)
pr_cont("%i ", i);
pr_cont("]\n");
I get:
6,173,0;[0 1 2 3 ]
And if I fill the cont buffer and forcefully hold the console sem
during all that, and we can't merge anymore, I get:
6,167,0;[
4,168,0;0
4,169,0;1
4,170,0;2
4,171,0;3
4,172,0;]
But the output is still all fine for both lines:
[ 0.000000] [0 1 2 3 ]
[ 0.000000] [0 1 2 3 ]
What do I miss?
Kay
next prev parent reply other threads:[~2012-07-09 22:10 UTC|newest]
Thread overview: 51+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-06-25 19:05 [PATCH v3] printk: Have printk() never buffer its data Steven Rostedt
2012-06-25 22:07 ` Andrew Morton
2012-06-25 23:55 ` Greg Kroah-Hartman
2012-06-26 0:01 ` Linus Torvalds
2012-06-26 0:23 ` Greg Kroah-Hartman
2012-06-26 0:40 ` Linus Torvalds
2012-06-26 0:56 ` Kay Sievers
2012-06-26 1:40 ` Linus Torvalds
2012-06-26 16:07 ` Kay Sievers
2012-06-26 16:30 ` Joe Perches
2012-06-26 16:58 ` Greg Kroah-Hartman
2012-06-26 17:00 ` Kay Sievers
2012-06-26 17:02 ` Greg Kroah-Hartman
2012-06-26 18:34 ` Greg Kroah-Hartman
2012-06-26 18:38 ` Greg Kroah-Hartman
2012-06-26 18:48 ` Greg Kroah-Hartman
2012-06-27 15:13 ` Steven Rostedt
2012-06-27 15:18 ` Steven Rostedt
2012-06-27 15:26 ` Kay Sievers
2012-06-28 7:38 ` Kay Sievers
2012-06-28 1:48 ` Greg Kroah-Hartman
2012-06-28 1:54 ` Steven Rostedt
2012-06-28 2:55 ` Steven Rostedt
2012-06-29 5:30 ` Greg Kroah-Hartman
2012-06-29 11:18 ` Steven Rostedt
2012-06-29 15:40 ` Greg Kroah-Hartman
2012-06-28 5:00 ` Joe Perches
2012-07-05 7:03 ` Michael Neuling
2012-07-05 8:39 ` Kay Sievers
2012-07-05 8:53 ` Kay Sievers
2012-07-05 10:20 ` Michael Neuling
2012-07-05 11:47 ` Kay Sievers
2012-07-05 12:50 ` Kay Sievers
2012-07-06 0:41 ` Michael Neuling
2012-07-06 0:56 ` Kay Sievers
2012-07-06 3:39 ` Michael Neuling
2012-07-06 3:47 ` Michael Neuling
2012-07-06 10:46 ` Kay Sievers
2012-07-06 15:12 ` Kay Sievers
2012-07-06 21:04 ` Michael Neuling
2012-07-08 17:55 ` Kay Sievers
2012-07-09 17:09 ` Greg Kroah-Hartman
2012-07-09 17:15 ` Joe Perches
2012-07-09 22:36 ` Michael Neuling
2012-07-09 21:42 ` Joe Perches
2012-07-09 22:10 ` Kay Sievers [this message]
2012-07-09 22:29 ` Joe Perches
2012-07-09 22:40 ` Kay Sievers
2012-07-09 23:32 ` Joe Perches
2012-07-09 23:41 ` Joe Perches
2012-06-26 0:18 ` Joe Perches
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAPXgP11Qsj9yGM5nSbRJ_YPFZcr2Km6w4R8wPGU1AytyQChYXQ@mail.gmail.com \
--to=kay@vrfy.org \
--cc=akpm@linux-foundation.org \
--cc=fengguang.wu@intel.com \
--cc=gregkh@linuxfoundation.org \
--cc=joe@perches.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linuxppc-dev@ozlabs.org \
--cc=mikey@neuling.org \
--cc=mingo@elte.hu \
--cc=paulmck@us.ibm.com \
--cc=rostedt@goodmis.org \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).