From owner-freebsd-hackers Mon Jan 3 15:25:56 2000 Delivered-To: freebsd-hackers@freebsd.org Received: from staff.accessus.net (staff.accessus.net [209.145.151.3]) by hub.freebsd.org (Postfix) with ESMTP id E811A1520E for ; Mon, 3 Jan 2000 15:25:41 -0800 (PST) (envelope-from doogie@staff.accessus.net) Received: by staff.accessus.net with Internet Mail Service (5.5.2650.21) id ; Mon, 3 Jan 2000 17:25:31 -0600 Message-ID: From: Jason Young To: 'Matthew Dillon' , George Cox Cc: freebsd-hackers@FreeBSD.ORG Subject: RE: Repeated softupdates panics in 3.3-STABLE Date: Mon, 3 Jan 2000 17:25:27 -0600 MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2650.21) Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01BF5641.D387C89C" Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01BF5641.D387C89C Content-Type: text/plain; charset="iso-8859-1" A "pending ops" panic can be induced in fairly short order by running the SMTP performance tests that come with Postfix. Specifically, run smtpstone/smtp-source running many parallel deliveries into a Postfix mail daemon setup running on the same machine. I had always assumed that this meant softupdates was getting too behind (the test delivers hundreds of mails per second with decent disk setups). I wasn't able to mentally parse the code well enough to confirm that. This is in all 3.x revisions that I've tested, since from back around the time of 3.0-CURRENT's great ELF transition, to 3.3-RELEASE. If somebody's interested, I can create this setup and try to duplicate the problem again. > -----Original Message----- > From: Matthew Dillon [mailto:dillon@apollo.backplane.com] > Sent: Monday, January 03, 2000 4:21 PM > To: George Cox > Cc: freebsd-hackers@FreeBSD.ORG > Subject: Re: Repeated softupdates panics in 3.3-STABLE > > > :On 03/01 16:29, Keith Stevenson wrote: > : > :> It looks like I may have spoken too soon when I mentioned > that I had no > :> problems with softupdates on my postfix based mail server. > I have now had > :> two panics in the last month with a panicstr of > "softdep_lock: locking > :> against myself". I thought that the first one might have > been a fluke until > :> it repeated itself today. > : > :I too have seen this "softdep_lock: locking against myself" > panic on a Postfix > :server. I was able to trigger it I think maybe twice by > issuing a 'postfix > :flush' command. :-/ This _was_ some months ago, when there > was the odd commit > :to the softupdates code going in, which suggests it's kind > of a long standing > :bug. :-( > > Well, in Keith's case the locking-against-myself panic is not the > cause, but the effect of the 'softdep_fsync: pending ops' panic > that occured just before it. > > I've never seeing a pending ops panic before, this is going to be > one for Kirk to track down. Be sure to keep your core dump and > your debug kernel. In fact, if you could gzip them both > up and make > them available to me and Kirk via a hidden ftp or hidden URL I > would appreciate it. NOTE! Kernel core dumps often contain > sensitive information such as pieces of the password file, do not > make your core available to the general lists! > > -Matt > Matthew Dillon > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-hackers" in the body of the message > ------_=_NextPart_001_01BF5641.D387C89C Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable RE: Repeated softupdates panics in 3.3-STABLE

A "pending ops" panic can be induced in = fairly short order by running the SMTP performance tests that come with = Postfix. Specifically, run smtpstone/smtp-source running many parallel = deliveries into a Postfix mail daemon setup running on the same = machine. I had always assumed that this meant softupdates was getting = too behind (the test delivers hundreds of mails per second with decent = disk setups). I wasn't able to mentally parse the code well enough to = confirm that.

This is in all 3.x revisions that I've tested, since = from back around the time of 3.0-CURRENT's great ELF transition, to = 3.3-RELEASE. If somebody's interested, I can create this setup and try = to duplicate the problem again.

> -----Original Message-----
> From: Matthew Dillon [mailto:dillon@apollo.backpla= ne.com]
> Sent: Monday, January 03, 2000 4:21 PM
> To: George Cox
> Cc: freebsd-hackers@FreeBSD.ORG
> Subject: Re: Repeated softupdates panics in = 3.3-STABLE
>
>
> :On 03/01 16:29, Keith Stevenson wrote:
> :
> :> It looks like I may have spoken too soon = when I mentioned
> that I had no
> :> problems with softupdates on my postfix = based mail server.
>  I have now had
> :> two panics in the last month with a = panicstr of
> "softdep_lock: locking
> :> against myself".  I thought = that the first one might have
> been a fluke until
> :> it repeated itself today.
> :
> :I too have seen this "softdep_lock: = locking against myself"
> panic on a Postfix
> :server.  I was able to trigger it I think = maybe twice by
> issuing a 'postfix
> :flush' command. :-/  This _was_ some = months ago, when there
> was the odd commit
> :to the softupdates code going in, which = suggests it's kind
> of a long standing
> :bug. :-(
>
>     Well, in Keith's case = the locking-against-myself panic is not the
>     cause, but the effect = of the 'softdep_fsync: pending ops' panic
>     that occured just = before it.
>
>     I've never seeing a = pending ops panic before, this is going to be
>     one for Kirk to track = down.  Be sure to keep your core dump and
>     your debug = kernel.  In fact, if you could gzip them both
> up and make
>     them available to me = and Kirk via a hidden ftp or hidden URL I
>     would appreciate = it.  NOTE!  Kernel core dumps often contain
>     sensitive information = such as pieces of the password file, do not
>     make your core = available to the general lists!
>
>       =         =         =         =         -Matt
>       =         =         =         =         Matthew Dillon
>       =         =         =         =         = <dillon@backplane.com>
>
>
> To Unsubscribe: send mail to = majordomo@FreeBSD.org
> with "unsubscribe freebsd-hackers" in = the body of the message
>

------_=_NextPart_001_01BF5641.D387C89C-- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message