NetBSD Problem Report #18789
Received: (qmail 21439 invoked by uid 605); 24 Oct 2002 10:53:40 -0000
Message-Id: <200210241041.g9OAfmB00498@k2.vaasje.org>
Date: Thu, 24 Oct 2002 12:41:48 +0200 (CEST)
From: fvdl@wasabisystems.com
Sender: gnats-bugs-owner@netbsd.org
Reply-To: fvdl@wasabisystems.com
To: gnats-bugs@gnats.netbsd.org
Subject: amap_lookups panic after uvm_map changes
X-Send-Pr-Version: 3.95
>Number: 18789
>Category: kern
>Synopsis: amap_lookups panic after uvm_map changes
>Confidential: no
>Severity: serious
>Priority: high
>Responsible: atatat
>State: closed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Thu Oct 24 10:54:00 +0000 2002
>Closed-Date: Thu Oct 24 20:50:27 +0000 2002
>Last-Modified: Thu Oct 24 20:50:27 +0000 2002
>Originator: Frank van der Linden
>Release: NetBSD 1.6J
>Organization:
poor
>Environment:
System: NetBSD k2 1.6J NetBSD 1.6J (K2) #1: Thu Oct 24 12:26:27 CEST 2002 fvdl@k2:/work/trees/current/sys/arch/i386/compile/K2 i386
Architecture: i386
Machine: i386
>Description:
A certain series of events makes the kernel panic with the following
trace (which is unfortunately not very interesting):
panic: amap_lookups: offset out of range
Stopped in pid 467 (java_vm) at cpu_Debugger+0x4: leave
db{1}>
db{1}> t
cpu_Debugger(e64ded18,8,3,0,e6361448) at cpu_Debugger+0x4
panic(c0333ea0,1,e64bee40,c0208138,e647e2c8) at panic+0xad
amap_lookups(e647e2f8,49000,e64beef4,8,e64ded18,c0334d80,2cf,0) at amap_lookups+0x51
uvm_fault(e636138c,4fd7c000,0,2,0) at uvm_fault+0x3b8
trap() at trap+0x5aa
--- trap (number 6) ---
0x4825d29d:
>How-To-Repeat:
Install the netscape7 pkg (Linux binary), including Java support.
Relevant packages:
suse_base-7.3 Linux compatibility package
suse_x11-7.3 Linux compatibility package for X11 binaries
suse_gtk-7.3 Linux compatibility package for GTK binaries
netscape7-0nb2 Netscape Browser/Mail- and Newsreader suite version 7
Start the browser.
Open http://news.bbc.co.uk/
When you get to the main page, watch the crash.
>Fix:
Reverting to uvm_map.c revision 1.120 prevents the panic.
>Release-Note:
>Audit-Trail:
From: enami tsugutomo <enami@sm.sony.co.jp>
To: fvdl@wasabisystems.com
Cc: gnats-bugs@gnats.netbsd.org, netbsd-bugs@netbsd.org
Subject: Re: kern/18789: amap_lookups panic after uvm_map changes
Date: 24 Oct 2002 20:34:03 +0900
fvdl@wasabisystems.com writes:
> panic: amap_lookups: offset out of range
:
> >Fix:
> Reverting to uvm_map.c revision 1.120 prevents the panic.
If the merge was bi-merge, and the prev_entry has amap, need to extend
amap again?
enami.
From: Andrew Brown <atatat@atatdot.net>
To: enami tsugutomo <enami@sm.sony.co.jp>
Cc: fvdl@wasabisystems.com, gnats-bugs@gnats.netbsd.org,
netbsd-bugs@netbsd.org
Subject: Re: kern/18789: amap_lookups panic after uvm_map changes
Date: Thu, 24 Oct 2002 09:01:10 -0400
>> panic: amap_lookups: offset out of range
> :
>> >Fix:
>> Reverting to uvm_map.c revision 1.120 prevents the panic.
>
>If the merge was bi-merge, and the prev_entry has amap, need to extend
>amap again?
bimerge only currently occurs in the kernel, where no amaps are
involved.
entries involving amaps aren't forward (or bi) merged, only possibly
back merged.
--
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org * "ah! i see you have the internet
twofsonet@graffiti.com (Andrew Brown) that goes *ping*!"
werdna@squooshy.com * "information is power -- share the wealth."
From: enami tsugutomo <enami@but-b.or.jp>
To: Andrew Brown <atatat@atatdot.net>
Cc: fvdl@wasabisystems.com, gnats-bugs@gnats.netbsd.org,
netbsd-bugs@netbsd.org
Subject: Re: kern/18789: amap_lookups panic after uvm_map changes
Date: Thu, 24 Oct 2002 22:42:33 +0900 (JST)
> bimerge only currently occurs in the kernel, where no amaps are
> involved.
Doesn't it occur theoretically, or just stat counter doesn't count up?
> entries involving amaps aren't forward (or bi) merged, only possibly
> back merged.
I can read the code that forwared merge won't occur if next entry has
amap, but can't find the code to inhibit bimerge when prev_entry has
amap.
enami.
Index: uvm_map.c
===================================================================
RCS file: /cvsroot/syssrc/sys/uvm/uvm_map.c,v
retrieving revision 1.121
diff -u -r1.121 uvm_map.c
--- uvm_map.c 2002/10/18 13:18:42 1.121
+++ uvm_map.c 2002/10/24 13:35:58
@@ -776,11 +776,10 @@
UVMCNT_DECR(map_kbackmerge);
UVMCNT_INCR(map_kbimerge);
} else {
- UVMCNT_DECR(map_kbackmerge);
- UVMCNT_INCR(map_kbimerge);
+ UVMCNT_DECR(map_ubackmerge);
+ UVMCNT_INCR(map_ubimerge);
}
- }
- else {
+ } else {
if (kmap)
UVMCNT_INCR(map_kforwmerge);
else
From: Andrew Brown <atatat@atatdot.net>
To: enami tsugutomo <enami@but-b.or.jp>
Cc: fvdl@wasabisystems.com, gnats-bugs@gnats.netbsd.org,
netbsd-bugs@netbsd.org
Subject: Re: kern/18789: amap_lookups panic after uvm_map changes
Date: Thu, 24 Oct 2002 10:31:53 -0400
>> bimerge only currently occurs in the kernel, where no amaps are
>> involved.
>
>Doesn't it occur theoretically, or just stat counter doesn't count up?
currently, since amaps are not extended *backwards*, there's no way to
merge a new entry forwards with an existing entry *if* the following
entry is backed by an amap. a bimerge is simply a new allocation that
filled the gap between two existing allocations. plain forward merges
actually occur rather infrequently.
since kernels don't use amaps, only object references to the kernel
object, forward merges (which are usually bimerges) can occur there.
regular applications typically don't make allocations that are
candidates for forward merges (ie, user space allocations usually
proceed in a forwards direction, are usually backed by amaps, and thus
only incur backmerges).
>> entries involving amaps aren't forward (or bi) merged, only possibly
>> back merged.
>
>I can read the code that forwared merge won't occur if next entry has
>amap, but can't find the code to inhibit bimerge when prev_entry has
>amap.
bimerge is simply a backmerge followed by a forward merge. the
backward merge probably occurs happily, and then the forward merge
fails because the following entry has an amap (if it has nothing
backing it, it will actually succeed, but this is exceedingly rare).
take a look at the stats i gathered over a period of three days at
http://www.graffiti.com/uvm.html
the userspace forward merges and forward merge failures indicated
there were all incurred by a program i specifically wrote to exercise
that code path. note that no bimerges occurred.
>Index: uvm_map.c
>===================================================================
>RCS file: /cvsroot/syssrc/sys/uvm/uvm_map.c,v
>retrieving revision 1.121
>diff -u -r1.121 uvm_map.c
>--- uvm_map.c 2002/10/18 13:18:42 1.121
>+++ uvm_map.c 2002/10/24 13:35:58
>@@ -776,11 +776,10 @@
> UVMCNT_DECR(map_kbackmerge);
> UVMCNT_INCR(map_kbimerge);
> } else {
>- UVMCNT_DECR(map_kbackmerge);
>- UVMCNT_INCR(map_kbimerge);
>+ UVMCNT_DECR(map_ubackmerge);
>+ UVMCNT_INCR(map_ubimerge);
> }
ah. good point. thanks.
>- }
>- else {
>+ } else {
> if (kmap)
> UVMCNT_INCR(map_kforwmerge);
> else
heh. style nits. cool.
--
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org * "ah! i see you have the internet
twofsonet@graffiti.com (Andrew Brown) that goes *ping*!"
werdna@squooshy.com * "information is power -- share the wealth."
From: enami tsugutomo <enami@but-b.or.jp>
To: Andrew Brown <atatat@atatdot.net>
Cc: fvdl@wasabisystems.com, gnats-bugs@gnats.netbsd.org,
netbsd-bugs@netbsd.org
Subject: Re: kern/18789: amap_lookups panic after uvm_map changes
Date: Thu, 24 Oct 2002 23:52:41 +0900 (JST)
> currently, since amaps are not extended *backwards*, there's no way to
> merge a new entry forwards with an existing entry *if* the following
> entry is backed by an amap. a bimerge is simply a new allocation that
> filled the gap between two existing allocations. plain forward merges
> actually occur rather infrequently.
Suppose following case:
prev_entry wi/ amap | gap will be filled by new entry | next_entry w/o amap
A B C
The resulting entry covers A+B+C but amap only covers A+B, doesn't it?
enami.
From: Andrew Brown <atatat@atatdot.net>
To: enami tsugutomo <enami@but-b.or.jp>
Cc: fvdl@wasabisystems.com, gnats-bugs@gnats.netbsd.org,
netbsd-bugs@netbsd.org
Subject: Re: kern/18789: amap_lookups panic after uvm_map changes
Date: Thu, 24 Oct 2002 11:40:48 -0400
>> currently, since amaps are not extended *backwards*, there's no way to
>> merge a new entry forwards with an existing entry *if* the following
>> entry is backed by an amap. a bimerge is simply a new allocation that
>> filled the gap between two existing allocations. plain forward merges
>> actually occur rather infrequently.
>
>Suppose following case:
>
>prev_entry wi/ amap | gap will be filled by new entry | next_entry w/o amap
>
> A B C
>
>The resulting entry covers A+B+C but amap only covers A+B, doesn't it?
an excellent idea! let me try it out...
--
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org * "ah! i see you have the internet
twofsonet@graffiti.com (Andrew Brown) that goes *ping*!"
werdna@squooshy.com * "information is power -- share the wealth."
From: Andrew Brown <atatat@atatdot.net>
To: enami tsugutomo <enami@but-b.or.jp>
Cc: fvdl@wasabisystems.com, gnats-bugs@gnats.netbsd.org,
netbsd-bugs@netbsd.org
Subject: Re: kern/18789: amap_lookups panic after uvm_map changes
Date: Thu, 24 Oct 2002 12:07:21 -0400
>>Suppose following case:
>>
>>prev_entry wi/ amap | gap will be filled by new entry | next_entry w/o amap
>>
>> A B C
>>
>>The resulting entry covers A+B+C but amap only covers A+B, doesn't it?
>
>an excellent idea! let me try it out...
wonderful. this program:
#define BASE ((void*)0x80000000)
#define PAGE 4096
int
main(int argc, char *argv[])
{
int *a, *b, *c;
/* allocate the "first" page */
a = mmap(BASE + 0 * PAGE, PAGE, PROT_READ|PROT_WRITE, MAP_ANON, -1, 0);
/* cause it to be backed by an amap */
a[0] = 1;
/* allocate the "last" page */
c = mmap(BASE + 2 * PAGE, PAGE, PROT_READ|PROT_WRITE, MAP_ANON, -1, 0);
/* allocate the "middle" page, which gets bimerged */
b = mmap(BASE + 1 * PAGE, PAGE, PROT_READ|PROT_WRITE, MAP_ANON, -1, 0);
/* die, you gravy sucking pig dog */
c[0] = 1;
}
will cause the same problem. patch forthcoming. thanks. :)
--
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org * "ah! i see you have the internet
twofsonet@graffiti.com (Andrew Brown) that goes *ping*!"
werdna@squooshy.com * "information is power -- share the wealth."
From: Andrew Brown <atatat@netbsd.org>
To: gnats-bugs@gnats.netbsd.org
Cc:
Subject: pr/18789 CVS commit: syssrc/sys/uvm
Date: Thu, 24 Oct 2002 23:38:01 +0300 (EEST)
Module Name: syssrc
Committed By: atatat
Date: Thu Oct 24 20:38:00 UTC 2002
Modified Files:
syssrc/sys/uvm: uvm_map.c
Log Message:
Call amap_extend() a second time in the case of a bimerge (both
backwards and forwards) if the previous entry was backed by an amap.
Fixes pr kern/18789, where netscape 7 + a java applet actually manage
to incur forward and bimerges in userspace.
Code reviewed by fvdl and thorpej.
To generate a diff of this commit:
cvs rdiff -r1.121 -r1.122 syssrc/sys/uvm/uvm_map.c
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
State-Changed-From-To: open->closed
State-Changed-By: atatat
State-Changed-When: Thu Oct 24 13:49:08 PDT 2002
State-Changed-Why:
the bug has been fixed, along with incorporating enami's suggested tweaks.
Responsible-Changed-From-To: kern-bug-people->atatat
Responsible-Changed-By: atatat
Responsible-Changed-When: Thu Oct 24 13:49:08 PDT 2002
Responsible-Changed-Why:
the problem was in my code, and i've fixed it (with enami's help, for which
i am very grateful).
>Unformatted:
(Contact us)
$NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.