INN with large file support dumps core under Solaris 2.7
Joe St Sauver
JOE at OREGON.UOREGON.EDU
Fri Feb 25 19:10:03 UTC 2000
Had a mysterious experience whereby today's snapshot would mysteriously
die with nothing to the syslog; trussing innd -d -f resulted in output
that looked like
....
brk(0x002E7000) = 0
ioctl(6, TCGETA, 0xDFFFF784) Err#25 ENOTTY
read(6, " # # $ R e v i s i o".., 8192) = 325
read(6, 0x002D900C, 8192) = 0
llseek(6, 0, SEEK_SET) = 0
read(6, " # # $ R e v i s i o".., 8192) = 325
read(6, 0x002D900C, 8192) = 0
llseek(6, 0, SEEK_CUR) = 325
close(6) = 0
open64("/dev/null", O_RDONLY) = 6
open64("/news/active", O_RDWR) = 16
fcntl(16, F_GETFD, 0x00000000) = 0
fcntl(16, F_SETFD, 0x00000001) = 0
fstat64(16, 0xDFFFF940) = 0
mmap64(0x00000000, 1186451, PROT_READ|PROT_WRITE, MAP_SHARED, 16, 0) = 0xDF1C0000
brk(0x004DE000) = 0
brk(0x004FF000) = 0
brk(0x00700000) = 0
brk(0x00901000) = 0
brk(0x0093D800) = 0
open64("/news/etc/newsfeeds", O_RDONLY) = 17
fstat64(17, 0xDFFFF940) = 0
read(17, " M E / n e w s . c g o c".., 3751) = 3751
close(17) = 0
Incurred fault #6, FLTBOUNDS %pc = 0x0005981C
siginfo: SIGSEGV SEGV_MAPERR addr=0xE640D77C
Received signal #11, SIGSEGV [default]
siginfo: SIGSEGV SEGV_MAPERR addr=0xE640D77C
*** process killed ***
Now obviously that's an odd argument to the read(17, ... call, but I am equally
suspicious of the mmap'd files based on past experience with inadequately
sized mmap'd regions. This is with inn-2000-02-25_03-01 under Solaris 2.7
on sun4d (SPARC) architecture built using
./configure --prefix=/news \
--bindir=/news/bin \
--mandir=/usr/local/man \
--with-perl \
--with-db-dir=/news \
--with-spool-dir=/var/spool/news \
--with-log-dir=/var/log/news \
--with-tmp-path=/news/tmp \
--with-sendmail=/usr/lib/sendmail \
--with-largefiles --disable-enable-tagged-hash
Regards,
Joe
More information about the inn-workers
mailing list