Bug 19978
| Summary: | /usr/bin/make segfaults w/o Makefile for root (over NFS) | ||
|---|---|---|---|
| Product: | Base System | Reporter: | pfeifer <pfeifer> |
| Component: | bin | Assignee: | Will Andrews <will> |
| Status: | Closed FIXED | ||
| Severity: | Affects Only Me | ||
| Priority: | Normal | ||
| Version: | 4.0-RELEASE | ||
| Hardware: | Any | ||
| OS: | Any | ||
|
Description
pfeifer
2000-07-17 12:20:01 UTC
Responsible Changed From-To: freebsd-bugs->will Over to MAINTAINER. State Changed From-To: open->closed Does not happen for me. Probably a NFS fluke with Solaris. On Sun, 8 Oct 2000 will@freebsd.org wrote: > Synopsis: /usr/bin/make segfaults w/o Makefile for root (over NFS) > > State-Changed-From-To: open->closed > State-Changed-By: will > State-Changed-When: Sun Oct 8 22:04:34 PDT 2000 > State-Changed-Why: > Does not happen for me. Probably a NFS fluke with Solaris. First of all, if it *was* a NFS fluke with Solaris, it still would be a bug. After all a FreeBSD program *is* crashing (ls for example, works as expected) and Solaris NFS *is* the standard everone else has to be measured again. Second, and more important: I just managed to reproduce this on an NFS mount from a 4.1-RELEASE client to a 4.1-RELEASE server (NFS v2, UDP). deneb[81]:/sw/FreeBSD% ls -lad 00TEST/ drwx------ 2 pfeifer sysop 512 Oct 9 09:57 00TEST/ deneb[82]:/sw/FreeBSD% cd 00TEST/ deneb[83]:/sw/FreeBSD/00TEST% su Password: deneb# make Segmentation fault Please reopen this PR. Gerald -- Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/ State Changed From-To: closed->open Reopened at submitters request. On Sun, 5 Nov 2000, David Malone wrote: > I've reopened the PR. Thanks! > Can you either get a back trace of the core from make using gdb or > ktrace it, so we can see what it was doing when it died? I've done both! ;-) This is on 4.1-RELEASE with make from the current CVS tree: #0 Dir_Init () at dir.c:231 231 dot->refCount += 1; (gdb) bt #0 Dir_Init () at dir.c:231 #1 0x8051e10 in main (argc=1, argv=0xbfbff964) at main.c:652 #2 0x8049745 in _start () and 15545 make CALL stat(0xbfbff4b0,0xbfbfebe4) 15545 make NAMI "obj.i386" 15545 make RET stat -1 errno 13 Permission denied 15545 make CALL stat(0x805c78f,0xbfbfebe4) 15545 make NAMI "obj" 15545 make RET stat -1 errno 13 Permission denied 15545 make CALL stat(0xbfbff4b0,0xbfbfebe4) 15545 make NAMI ".../x" ---- THIS IS THE UNREADABLE DIRECTORY! ---- 15545 make RET stat -1 errno 2 No such file or directory 15545 make CALL readlink(0x280f4154,0xbfbfeb74,0x3f) 15545 make NAMI "/etc/malloc.conf" 15545 make RET readlink -1 errno 2 No such file or directory 15545 make CALL mmap(0,0x1000,0x3,0x1002,0xffffffff,0,0,0) 15545 make RET mmap 672186368/0x2810c000 15545 make CALL break(0x8062000) 15545 make RET break 0 15545 make CALL break(0x8063000) 15545 make RET break 0 15545 make CALL stat(0x805b887,0xbfbfeb64) 15545 make NAMI "." 15545 make RET stat 0 15545 make CALL open(0x805b887,0x4,0x280f8190) 15545 make NAMI "." 15545 make RET open -1 errno 13 Permission denied 15545 make PSIG SIGSEGV SIG_DFL 15545 make NAMI "/tmp/make.core" Gerald -- Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/ On Sun, 5 Nov 2000, Gerald Pfeifer wrote: > On Sun, 5 Nov 2000, David Malone wrote: > > I've reopened the PR. > > Thanks! > > > Can you either get a back trace of the core from make using gdb or > > ktrace it, so we can see what it was doing when it died? > 15545 make CALL stat(0x805b887,0xbfbfeb64) > 15545 make NAMI "." > 15545 make RET stat 0 > 15545 make CALL open(0x805b887,0x4,0x280f8190) > 15545 make NAMI "." > 15545 make RET open -1 errno 13 Permission denied > 15545 make PSIG SIGSEGV SIG_DFL > 15545 make NAMI "/tmp/make.core" This output made it easy to reproduce the bug in a simpler way: $ cd /tmp $ mkdir z $ cd z $ chmod 0 /tmp/z $ make Segmentation fault Bruce Could the attached patch do something about it? It's against RELENG_4
(rev 1.10 of dir.c).
G'luck,
Peter
--
If this sentence were in Chinese, it would say something else.
On Tue, Nov 07, 2000 at 09:29:39PM +1100, Bruce Evans wrote:
> On Sun, 5 Nov 2000, Gerald Pfeifer wrote:
>
> > On Sun, 5 Nov 2000, David Malone wrote:
> > > I've reopened the PR.
> >
> > Thanks!
> >
> > > Can you either get a back trace of the core from make using gdb or
> > > ktrace it, so we can see what it was doing when it died?
>
> > 15545 make CALL stat(0x805b887,0xbfbfeb64)
> > 15545 make NAMI "."
> > 15545 make RET stat 0
> > 15545 make CALL open(0x805b887,0x4,0x280f8190)
> > 15545 make NAMI "."
> > 15545 make RET open -1 errno 13 Permission denied
> > 15545 make PSIG SIGSEGV SIG_DFL
> > 15545 make NAMI "/tmp/make.core"
>
> This output made it easy to reproduce the bug in a simpler way:
>
> $ cd /tmp
> $ mkdir z
> $ cd z
> $ chmod 0 /tmp/z
> $ make
> Segmentation fault
diff -urN src/usr.bin/make/dir.c mysrc/usr.bin/make/dir.c
--- src/usr.bin/make/dir.c Sat Sep 11 16:08:01 1999
+++ mysrc/usr.bin/make/dir.c Tue Nov 7 12:42:20 2000
@@ -225,6 +225,8 @@
*/
Dir_AddDir (openDirectories, ".");
dot = (Path *) Lst_DeQueue (openDirectories);
+ if (dot == (Path *) NIL)
+ err(1, "cannot open current directory");
/*
* We always need to have dot around, so we increment its reference count
On Tue, Nov 07, 2000 at 03:10:05AM -0800, Bruce Evans wrote:
> This output made it easy to reproduce the bug in a simpler way:
>
> $ cd /tmp
> $ mkdir z
> $ cd z
> $ chmod 0 /tmp/z
> $ make
> Segmentation fault
Eek! Shoo, bug! Shoo!
--
wca
What happened to this PR? Please note that it also contains a patch which fixes the problem! Gerald -- Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/ State Changed From-To: open->suspended Fixed; will be MFC'd to RELENG_[43] in 3 days (along with the bugfix for readdir() assumption, which has been in -current for a lot longer). State Changed From-To: suspended->closed MFC'd to RELENG_4 and RELENG_3 yesterday. |
