John Baldwin wrote:
On Thursday 22 February 2007 12:17, Kostik Belousov wrote:
On Tue, Feb 20, 2007 at 11:21:59AM -0500, John Baldwin wrote:
On Tuesday 20 February 2007 09:49, Kostik Belousov wrote:
On Mon, Feb 19, 2007 at 06:39:58PM -0500, John Baldwin wrote:
On Monday 19 February 2007 05:56, Konstantin Belousov wrote:
kib 2007-02-19 10:56:10 UTC
FreeBSD src repository
Modified files:
sys/kern vfs_syscalls.c
Log:
Remove union_dircheckp hook, it is not needed by new unionfs code anymore.
As consequence, getdirentries() no longer needs to drop/reacquire
directory vnode lock, that would allow it to be reclaimed in between.
I think there is at least one more copy of getdirentries() under sys/compat,
possibly multiple. Are you going to fix this in all of them?
Could you, please, point out where is it ? My grep told me that I removed
all references to the union_dircheckp, and I think that getdirentries() code
had to have references to it. My change modified two syscalls: getdirentries()
and ogetdirentries().
Well, I did find at least 3 ABIs (Linux, svr4, and ibcs2) that do their own
wrapper around VOP_READDIR(), but none of them had the union check in them.
As result, "mount -o union" does not work for compat binaries.
Do you know why the ABIs use cookies for VOP_READDIR() but getdirentries()
doesn't?
No, and it seems that cookies, as well as vfs_syscalls.c avoidance of cookies
(together with union_dircheckp) go back to at least 1994/1997 years, and
Lite-2 import (see commit logs for rev. 1.9 and 1.54 of vfs_syscalls.c).
What would be nice would be to have a kern_getdirentries() that took a
function pointer that took a directory record and did the actual
conversion + UIO so we could cut down on the duplicated code.
Actually, Linux does this, and it's a mess IMHO. The cure is much worse
than the disease.
Scott
_______________________________________________
cvs-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/cvs-all
To unsubscribe, send any mail to "[EMAIL PROTECTED]"
