Hello,
We use openssl im our application servers and encounter the following
problem on SunOS Solaris 5.9 on SPARC:
the used call
r = BIO_gets(slnpSSL[i].io, buf, maxlen);
sslError = SSL_get_error(slnpSSL[i].ssl, r);
switch (sslError) {
case SSL_ERROR_NONE:
len = r;
break;
default:
fprintf(stderr, "SSL_get_error: %d\n", sslError);
berr_exit("SSL read problem");
}
will fail with SSL_ERROR_WANT_READ if the read(2) call on the socket is
interrupted by SIGALRM; a truss show this like this:
read(8, 0x0032842B, 5) (sleeping...)
Received signal #14, SIGALRM, in read() [caught]
siginfo: SIG#0
read(8, 0x0032842B, 5) Err#4 EINTR
lwp_sigmask(SIG_SETMASK, 0x00000000, 0x00000000) = 0xFFBFFEFF
[0x0000FFFF]
time() = 1396594526
sigaction(SIGALRM, 0xFFBF5AE0, 0xFFBF5B80) = 0
alarm(60) = 0
setcontext(0xFFBF5F68)
fstat64(2, 0xFFBF5CF8) = 0
write(2, " S S L r e a d p r o".., 17) = 17
_exit(0)
i.e. the read(2) is not setup again by the system;
on a Linux system it is; the same code and situation look like this:
19595 10:58:24.757535 alarm(60) = 0
19595 10:58:24.757588 sigreturn() = 3
19595 10:58:24.757638 read(7, 0x810469b, 5) = ? ERESTARTSYS (To be restarted)
19595 10:59:24.757784 --- SIGALRM (Alarm clock) @ 0 (0) ---
19595 10:59:24.757966 time(NULL) = 1396601964
19595 10:59:24.758040 rt_sigaction(SIGALRM, {0x10000000080582b4, [], 0},
{0x10000000080582b4, [], 0}, 8) = 0
19595 10:59:24.758117 alarm(60) = 0
19595 10:59:24.758172 sigreturn() = 3
19595 10:59:24.758224 read(7, 0x810469b, 5) ...
Why is this different on Solaris and Linux?
matthias
--
Matthias Apitz | /"\ ASCII Ribbon Campaign: www.asciiribbon.org
E-mail: g...@unixarea.de | \ / - No HTML/RTF in E-mail
WWW: http://www.unixarea.de/ | X - No proprietary attachments
phone: +49-170-4527211 | / \ - Respect for open standards
______________________________________________________________________
OpenSSL Project http://www.openssl.org
User Support Mailing List openssl-users@openssl.org
Automated List Manager majord...@openssl.org
