Date: Mon, 26 Jun 1995 22:50:06 +0800 (WST) From: Peter Wemm <peter@haywire.DIALix.COM> To: current@freebsd.org Subject: bug in Linux^H^H^H^H^HDoom Emulator Message-ID: <Pine.SV4.3.91.950626224220.12512A-100000@haywire.DIALix.COM>
next in thread | raw e-mail | index | archive | help
Thanks for committing it Soren (?? is that how you write it in 7-bit ascii?).
If you compile a kernel with options DIAGNOSTIC, it running doom will
cause a panic inside ufs_access - the vnode in the VOP_ACCESS call in
linux_misc.c is not locked at this point.
I've had a look, and based on what I saw in kern_exec.c, I suggest that
the following patch *might* be in order... I dont know if it's correct
and have not yet compiled a kernel to test it, so caveat emptor!
kern_exec.c approximately does this:
NDINIT(... FOLLOW | LOCKLEAF ...)
namei(&nd).
vp = nd.....
[...]
err = VOP_ACCESS(vp)
err = VOP_OPEN(vp)
VOP_UNLOCK(vp)
err = vm_mmap(vp...)
I added LOCKLEAF and a VOP_UNLOCK() to i386/linux/linux_misc.c - I'd
appreciate knowing if it looks right or if it's going to cause the file
system to get eaten.. :-)
Cheers,
-Peter
(PS: Thanks for comitting it Soren! :-)
*** linux_misc.c.dist Mon Jun 26 01:32:37 1995
--- linux_misc.c Mon Jun 26 22:38:41 1995
***************
*** 188,194 ****
printf("Linux-emul(%d): uselib(%s)\n", p->p_pid, path);
#endif
! NDINIT(&ni, LOOKUP, FOLLOW, UIO_SYSSPACE, path, p);
if (error = namei(&ni))
return error;
--- 188,194 ----
printf("Linux-emul(%d): uselib(%s)\n", p->p_pid, path);
#endif
! NDINIT(&ni, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, path, p);
if (error = namei(&ni))
return error;
***************
*** 215,220 ****
--- 215,222 ----
if (error = VOP_OPEN(vnodep, FREAD, p->p_ucred, p))
return error;
+
+ VOP_UNLOCK(vnodep); /* lock no longer needed */
error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, 1024,
VM_PROT_READ, VM_PROT_READ, 0, (caddr_t)vnodep, 0);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.SV4.3.91.950626224220.12512A-100000>