Chris Gerhard wrote:
Alan Wright wrote:
If the engine has been restarted, the socket should get closed
or return an error on the read. A zero length read isn't
necessarily a problem, and the vscan code is retrying the read
because it is expecting more data.
While vscand is expecting more data it is never going to arrive down
that fd as a read returning zero bytes signifies EOF and will never get
any more data. If the stream was marked for non blocking IO it would
have returned -1 with errno set otherwise it will block until there is
data or return 0 if there has been a hangup, ie the other end of the
stream is closed.
Yep, feel free to file the bug.
Don't worry about the network capture.
Alan
If you file a bug, please attach a network capture showing the
interaction when the problem occurs.
ok.
Thanks,
Alan
Chris Gerhard wrote:
My home server's vscand just got itself into a loop consuming a CPU
while read returned 0 bytes.
A bit of D got the stack where it was doing this:
0 3910 read:return
libc.so.1`__read+0x7
vscand`vs_icap_read+0x25
vscand`vs_icap_send_chunk+0x3e
vscand`vs_icap_respmod_request+0xcf
vscand`vs_icap_scan_file+0x184
vscand`vs_svc_scan_file+0xd1
vscand`vs_svc_async_scan+0x22
libc.so.1`_thrp_setup+0x9b
libc.so.1`_lwp_start
0 3910 read:return
libc.so.1`__read+0x7
vscand`vs_icap_read+0x25
vscand`vs_icap_send_chunk+0x3e
vscand`vs_icap_respmod_request+0xcf
vscand`vs_icap_scan_file+0x184
vscand`vs_svc_scan_file+0xd1
vscand`vs_svc_async_scan+0x22
libc.so.1`_thrp_setup+0x9b
libc.so.1`_lwp_start
Unfortunately I did not get the name of the file is was reading
(although I suspect it was a network connection).
The source confirms that it can't handle a return of 0 bytes:
1173 vs_icap_read(int fd, char *buf, int len)
1174 {
1175 char *ptr = buf;
1176 int resid = len;
1177 int bytes_read = 0;
1178
1179 while (resid > 0) {
1180 errno = 0;
1181 bytes_read = read(fd, ptr, resid);
1182 if (bytes_read < 0) {
1183 if (errno == EINTR)
1184 continue;
1185 else
1186 return (-1);
1187 }
1188 resid -= bytes_read;
1189 ptr += bytes_read;
1190 }
1191
1192 return (len - resid);
1193 }
I suspect the problem coincides with a virus engine being restarted.
I'll file a bug next week when I'm back in the office.
_______________________________________________
cifs-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/cifs-discuss
