Further to the last post, I've determined (lines preceded by '-' are debug
messages in my kernel):
*** sigtoproc sets the signal mask
- 3000BA224EC = 00020000 00000000 00000000 00000000
*** issig_forreal called which calls fsig which checks the masks and finds
signal 18 ready to send
- pid: 100454 sigbits[0] is zero - ssp: 2a100a19830
- pid: 100454 sigbits[1] is zero - ssp: 2a100a19830
- pid: 100454 temp.__sigbits[0]=131072 NBBY=8 lowbit(temp.__sigbits[0])=18
ssp: 3000ba224ec
*** sigdqe dequeues the signal and clears the mask
*** sshd used the signal() API which gets transformed to a sigaction with the
SA_RESETHAND flag set, so setsigact clears the handler and mask
*** sendsig32 prepares process to call the handler
- context setup to go to handler for 100454
*** sshd is currently in a critical section so the signal gets deferred until
do_exit_critical() calls take_deferred_signal() which drives a sigresend
syscall
- pid: 100454 sigresend - sig: 18 signinfo: 0 set: 7d502738
- - sigbits[0]: 0 sigbits[1]: 0
*** However, issig_forreal() and fsig() find that the mask zeroed so there's no
signal to be sent!!!
- pid: 100454 sigbits[0] is zero - ssp: 2a100a19830
- pid: 100454 sigbits[1] is zero - ssp: 2a100a19830
- pid: 100454 sigbits[0] is zero - ssp: 3000ba224ec
- pid: 100454 sigbits[1] is zero - ssp: 3000ba224ec
- issig(0)=0
Now sshd could be changed to use sigaction without the SA_RESETHAND flag, but
this hasn't been a problem for anyone else so I'm wondering if the resend logic
is complete.
Neale
This message posted from opensolaris.org
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
