On Mon, Jun 23, 2014 at 11:42:14PM +0100, Peter Maydell wrote: > On 23 June 2014 23:36, Paul Burton <p...@archlinuxmips.org> wrote: > > Actually no, I don't think you're right about that afterall. The > > argument union itself is never modified. I imagine if it were then it > > would be painful in the case of the semctl syscall where the union is > > passed directly as an argument, rather than as a pointer as it is for > > the ipc syscall. > > > > What may be modified is the data pointed to by the pointers within union > > semun. That is already handled by do_semctl & the translate functions it > > calls. > > Except if you look at do_semctl you see code like: > case GETVAL: > case SETVAL: > arg.val = tswap32(target_su.val); > ret = get_errno(semctl(semid, semnum, cmd, arg)); > target_su.val = tswap32(arg.val); > break; > > which clearly is just modifying fields in the target_semun union. > So something's wrong (probably that code)...
Yes, both Linux & man semctl agree that GETVAL returns the value of
the semaphore as the return value of the syscall. So I believe the
assignment to target_su.val there is (functionally harmless) garbage.
Thanks,
Paul
signature.asc
Description: Digital signature
