| Message ID | 20250224172337.2009871-1-csander@purestorage.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | io_uring/waitid: remove #ifdef CONFIG_COMPAT | expand |
On 2/24/25 10:23 AM, Caleb Sander Mateos wrote: > io_is_compat() is already defined to return false if CONFIG_COMPAT is > disabled. So remove the additional #ifdef CONFIG_COMPAT guards. Let the > compiler optimize out the dead code when CONFIG_COMPAT is disabled. Would you mind if I fold this into Pavel's patch? I can keep it standalone too, just let me know.
On Mon, Feb 24, 2025 at 9:44 AM Jens Axboe <axboe@kernel.dk> wrote: > > On 2/24/25 10:23 AM, Caleb Sander Mateos wrote: > > io_is_compat() is already defined to return false if CONFIG_COMPAT is > > disabled. So remove the additional #ifdef CONFIG_COMPAT guards. Let the > > compiler optimize out the dead code when CONFIG_COMPAT is disabled. > > Would you mind if I fold this into Pavel's patch? I can keep it > standalone too, just let me know. Fine by me, though I thought Pavel was suggesting keeping it separate: https://lore.kernel.org/io-uring/da109d01-7aab-4205-bbb1-f5f1387f1847@gmail.com/T/#u Thanks, Caleb
On 2/24/25 10:53 AM, Caleb Sander Mateos wrote: > On Mon, Feb 24, 2025 at 9:44 AM Jens Axboe <axboe@kernel.dk> wrote: >> >> On 2/24/25 10:23 AM, Caleb Sander Mateos wrote: >>> io_is_compat() is already defined to return false if CONFIG_COMPAT is >>> disabled. So remove the additional #ifdef CONFIG_COMPAT guards. Let the >>> compiler optimize out the dead code when CONFIG_COMPAT is disabled. >> >> Would you mind if I fold this into Pavel's patch? I can keep it >> standalone too, just let me know. > > Fine by me, though I thought Pavel was suggesting keeping it separate: > https://lore.kernel.org/io-uring/da109d01-7aab-4205-bbb1-f5f1387f1847@gmail.com/T/#u I'm reading it as he has other stuff that will go on top. I don't see any reason to double stage this part, might as well remove the CONFIG dependency at the same time, if it's doable. Pavel?
On 2/24/25 12:10 PM, Pavel Begunkov wrote: > On 2/24/25 17:55, Jens Axboe wrote: >> On 2/24/25 10:53 AM, Caleb Sander Mateos wrote: >>> On Mon, Feb 24, 2025 at 9:44?AM Jens Axboe <axboe@kernel.dk> wrote: >>>> >>>> On 2/24/25 10:23 AM, Caleb Sander Mateos wrote: >>>>> io_is_compat() is already defined to return false if CONFIG_COMPAT is >>>>> disabled. So remove the additional #ifdef CONFIG_COMPAT guards. Let the >>>>> compiler optimize out the dead code when CONFIG_COMPAT is disabled. >>>> >>>> Would you mind if I fold this into Pavel's patch? I can keep it >>>> standalone too, just let me know. >>> >>> Fine by me, though I thought Pavel was suggesting keeping it separate: >>> https://lore.kernel.org/io-uring/da109d01-7aab-4205-bbb1-f5f1387f1847@gmail.com/T/#u >> >> I'm reading it as he has other stuff that will go on top. I don't see >> any reason to double stage this part, might as well remove the >> CONFIG dependency at the same time, if it's doable. >> >> Pavel? > > I'm not sure why you'd want that, but I don't mind Just because it imho should've been in the initial commit, so I'd consider it more of a fixup commit. But if you're fine with it, I'll fold it in with a note.
On 2/24/25 17:55, Jens Axboe wrote: > On 2/24/25 10:53 AM, Caleb Sander Mateos wrote: >> On Mon, Feb 24, 2025 at 9:44 AM Jens Axboe <axboe@kernel.dk> wrote: >>> >>> On 2/24/25 10:23 AM, Caleb Sander Mateos wrote: >>>> io_is_compat() is already defined to return false if CONFIG_COMPAT is >>>> disabled. So remove the additional #ifdef CONFIG_COMPAT guards. Let the >>>> compiler optimize out the dead code when CONFIG_COMPAT is disabled. >>> >>> Would you mind if I fold this into Pavel's patch? I can keep it >>> standalone too, just let me know. >> >> Fine by me, though I thought Pavel was suggesting keeping it separate: >> https://lore.kernel.org/io-uring/da109d01-7aab-4205-bbb1-f5f1387f1847@gmail.com/T/#u > > I'm reading it as he has other stuff that will go on top. I don't see > any reason to double stage this part, might as well remove the > CONFIG dependency at the same time, if it's doable. > > Pavel? I'm not sure why you'd want that, but I don't mind
On Mon, 24 Feb 2025 10:23:36 -0700, Caleb Sander Mateos wrote: > io_is_compat() is already defined to return false if CONFIG_COMPAT is > disabled. So remove the additional #ifdef CONFIG_COMPAT guards. Let the > compiler optimize out the dead code when CONFIG_COMPAT is disabled. > > Applied, thanks! [1/1] io_uring/waitid: remove #ifdef CONFIG_COMPAT commit: 0cd64345c4ba127d27fa07a133d108ea92d38361 Best regards,
diff --git a/io_uring/waitid.c b/io_uring/waitid.c index 4034b7e3026f..54e69984cd8a 100644 --- a/io_uring/waitid.c +++ b/io_uring/waitid.c @@ -40,11 +40,10 @@ static void io_waitid_free(struct io_kiocb *req) kfree(req->async_data); req->async_data = NULL; req->flags &= ~REQ_F_ASYNC_DATA; } -#ifdef CONFIG_COMPAT static bool io_waitid_compat_copy_si(struct io_waitid *iw, int signo) { struct compat_siginfo __user *infop; bool ret; @@ -65,24 +64,21 @@ static bool io_waitid_compat_copy_si(struct io_waitid *iw, int signo) return ret; Efault: ret = false; goto done; } -#endif static bool io_waitid_copy_si(struct io_kiocb *req, int signo) { struct io_waitid *iw = io_kiocb_to_cmd(req, struct io_waitid); bool ret; if (!iw->infop) return true; -#ifdef CONFIG_COMPAT if (io_is_compat(req->ctx)) return io_waitid_compat_copy_si(iw, signo); -#endif if (!user_write_access_begin(iw->infop, sizeof(*iw->infop))) return false; unsafe_put_user(signo, &iw->infop->si_signo, Efault);
io_is_compat() is already defined to return false if CONFIG_COMPAT is disabled. So remove the additional #ifdef CONFIG_COMPAT guards. Let the compiler optimize out the dead code when CONFIG_COMPAT is disabled. Signed-off-by: Caleb Sander Mateos <csander@purestorage.com> --- io_uring/waitid.c | 4 ---- 1 file changed, 4 deletions(-)