Are these overflows a security issue? Like giving up valid kernel memory.
-- Cryo:William J. Coldwell ARIN:WC25/AS7769 PGP:0xF97CC215/0x5E9944455 Warped, Inc. warped.com Founder/CTO 661-WARPED1 @warped @deadjournal NetBSD netbsd.org/pkgsrc.org President,Project Security,Social Media "Put on 3D glasses, otherwise you only see in 1½D.” [self opinion]; As mentioned in the other thread here, I am trying to make ktrace output working for netbsd32 emulated ktrace/kdump.
The patch below allows emulations to override the "msghdr" ktrace data output by various variants of sendmsg/recvmsg. This is simmply done by passing two new args to do_sys_sendmsg()/do_sys_recvmsg() and friends which provide alternative data to record in the ktrace record.
Since a full msgheader is not always conveniently available, it is also possible to pass a NULL pointer and ~0U as size to skip this ktrace record completely.
If passing NULL/0 as the new args, the native msghdr is output, which means most callers simply needed adding "NULL, 0," in the argument list.
A few compat versions are missing here, but will be added before commit.
Any objections?
Martin
<ktr_recvsend.patch>
|