Messages in this thread | | | Date | Wed, 27 Aug 2014 19:02:33 +0200 | From | Oleg Nesterov <> | Subject | Re: [PATCH 2/5] x86, fpu: don't drop_fpu() in __restore_xstate_sig() if use_eager_fpu() |
| |
I'm afraid you all already hate me, but let me continue to spam your mailboxes.
On 08/25, Oleg Nesterov wrote: > > I'll try to play with copy_from_user_in_atomic(), if nothing else just > to complete the discussion and see how the code can look in this case.
OK, to complete the discussion, the code looks simple unless I missed something,
if (ia32_fxstate) { /* * For 32-bit frames with fxstate, copy the user state to the * thread's fpu state, reconstruct fxstate from the fsave * header. Sanitize the copied state etc. */ struct xsave_struct *xsave = &tsk->thread.fpu.state->xsave; struct user_i387_ia32_struct env; bool done;
if (__copy_from_user(&env, buf, sizeof(env)) return -1;
for (done = false; !done; ) { if (fatal_signal_pending(current) || fault_in_pages_readable(buf_fx, state_size)) return -1;
preempt_disable(); pagefault_disable(); /* not really needed */ done = !__copy_from_user_inatomic(xsave, buf_fx, state_size); pagefault_enable(); if (likely(done)) { sanitize_restored_xstate(tsk, &env, xstate_bv, fx_only); if (__thread_has_fpu(tsk)) math_state_restore(); } else { fpu_finit(&tsk->thread.fpu); } preempt_enable(); }
return 0; }
and in some sense it is even simpler because we do not care about use_eager_fpu() or set/clear_used_math().
Does it look better than switch_fpu_xstate() hack?
However, this code can race with kernel_fpu_begin() if use_eager_fpu(). I _think_ that kernel_fpu_begin/end and irq_fpu_usable() need cleanups too and in any case. Will try to do, but I am not sure.
And to spam you even more, I'll send you a couple of other, more simple cleanups.
Oleg.
| |