Consistent Solaris crashes

[zellster]

Hello,

I am running lsws 1.5.7 Standard Edition on a Solaris 5.8 SPARC box (uname info: SunOS mach_name 5.8 Generic_108528-23 sun4u sparc SUNW,UltraAX-i2).

I am consistently seeing the following problem in the lsws process:

Received signal #14, SIGALRM, in poll() [caught]
poll(0x00229098, 5, 1000) Err#4 EINTR
setcontext(0xFFBEF728)
time() = 1094165817
poll(0x00229098, 5, 1000) = 1
accept(9, 0x001F0AFC, 0xFFBEF934, 1) = 12
accept(9, 0x001F0B1C, 0xFFBEF934, 1) Err#11 EAGAIN
fcntl(12, F_SETFD, 0x00000001) = 0
fstat64(12, 0xFFBEF750) = 0
getsockopt(12, 65535, 8192, 0xFFBEF850, 0xFFBEF84C, 2198285) = 0
setsockopt(12, 65535, 8192, 0xFFBEF850, 4, 2198285) = 0
fcntl(12, F_SETFL, 0x00000080) = 0
poll(0x00229098, 6, 1000) = 1
read(12, " G E T / l i b / f p b".., 2044) = 341
Incurred fault #5, FLTACCESS %pc = 0x00059FE4
siginfo: SIGBUS BUS_ADRALN addr=0x00218E03
Received signal #10, SIGBUS [default]
siginfo: SIGBUS BUS_ADRALN addr=0x00218E03
*** process killed ***

Any ideas what this could be? Any wonkish code performing word comparisons on unaligned data?

[mistwang]

Thank you for the bug report.

Yes, looks like it is. SPARC is picky about alignments. ;-)

Can you please send us the core file or the back trace of call stack? We have not been able to reproduce it in our lab.

Core file should be under /tmp/lshttpd/ directory.
You may have to use "coreadm" command to enable core dump for a setuid process, or just start the server process with the non-privilege user specified during installation.

Thank you very much!
George Wang

[zellster]

Hello,

Here's a stack trace. I will attach the core file in the next post.

# dbx /usr/local/lsws/bin/lshttpd.1.5.7 core
For information about new features see `help changes'
To remove this message, put `dbxenv suppress_startup_message 7.1' in your .dbxrc
Reading lshttpd.1.5.7
core file header read successfully
Reading ld.so.1
Reading libsocket.so.1
Reading libnsl.so.1
Reading librt.so.1
Reading libm.so.1
Reading libc.so.1
Reading libresolv.so.2
Reading libdl.so.1
Reading libmp.so.2
Reading libaio.so.1
Reading libc_psr.so.1
Reading nss_files.so.1
program terminated by signal BUS (invalid address alignment)
Symbol *0x83f0d0
dbx: duplicate type definition (0,1), assuming (int {assumed}), sclass 28: /usr/local/lsws/bin/lshttpd.1.5.7:/home/gwang/crossrel/release/httpd/httpd/main.cpp stab #30 int:t(0,1)=r(0,1);0020000000000;0017777777777;
(dbx) where
=>[1] newKeyValueBuf__7HttpReqRi(0x2164c4, 0x2165b8, 0x7e7e7e00, 0x76696100, 0x76696000, 0x74000000), at 0x59fe4
[2] newUnknownHeader__7HttpReq(0x2164c4, 0x216bd9, 0x59c00, 0xa, 0x216bd2, 0xa), at 0x5e9f8
[3] processHeaderLines__7HttpReq(0x2164c4, 0x216bd2, 0x216c8f, 0x3a, 0x21cd4, 0x4e4e0), at 0x59064
[4] processRequestLine__7HttpReq(0x2164c4, 0x0, 0x216ba4, 0x216c8f, 0x0, 0x0), at 0x58bc4
[5] processHeader__7HttpReq(0x2164c4, 0x2164c4, 0x7fc, 0x0, 0x0, 0x0), at 0x579cc
[6] readToHeaderBuf__14HttpConnection(0x21648c, 0x1, 0x1, 0x0, 0x0, 0x0), at 0x60598
[7] onReadEx__14HttpConnection(0x21648c, 0x61c14, 0x0, 0x0, 0x0, 0x0), at 0x61d44
[8] onRead__10HttpIOLinkP10HttpIOLink(0x21648c, 0x4df24, 0x0, 0x0, 0x0, 0x0), at 0x4df50
[9] handleEvents__10HttpIOLinks(0x21648c, 0x1, 0x4cecc, 0x0, 0x0, 0x0), at 0x4d02c
[10] processAllEvents__13PollfdReactor(0x205910, 0x205910, 0x3e8, 0x0, 0x20c44, 0x918f4), at 0x91d9c
[11] waitAndProcessEvents__6Polleri(0x205908, 0x3e8, 0x918b8, 0x0, 0x0, 0x0), at 0x9192c
[12] run__15EventDispatcher(0x227444, 0x227444, 0x0, 0x0, 0x0, 0x0), at 0x46644
[13] start__14HttpServerImpl(0x22742c, 0x178400, 0x179000, 0xffffffff, 0xfffffff8, 0x227d4d), at 0x13d4c
[14] start__10HttpServer(0x1eebe4, 0x1, 0xffbefd04, 0x0, 0x0, 0xff21f854), at 0x18e5c
[15] main__10HttpServeriPPc(0x1eebe4, 0x1, 0xffbefd04, 0x300, 0x225b4, 0xff19bc08), at 0x19270
[16] main(0x1, 0xffbefd04, 0xffbefd0c, 0x1eebc4, 0x0, 0x0), at 0x136cc
(dbx) quit

[zellster]

Link to core file:
http://zell.best.vwh.net/~zell/lsws.core

[mistwang]

Thank you very much!
It is fixed. The link to the new package is in the PM.

Thanks again. :-)
George Wang

[zellster]

Hello,

I have been running the updated Solaris version for the past six hours without a crash. Thanks!