Created attachment 264053 [details] Proposed change. The vt_mode::frsig signal is described to be unimplemented [1] in the vt driver and it does not seem to be used anywhere but we fail the VT_SETMODE call with EINVAL if it is not a valid signal number [2]. My understanding is that frsig is a historical artifact and that most modern *NIXes do not implement it (e.g. Linux [3] and Illumos [4]). If that is the case and there is no outlook for it to be ever implemented, we should probably not demand it. [1] https://github.com/freebsd/freebsd-src/blob/7cd7b849b0629a259f4034514cf0e47e82efc44a/sys/sys/consio.h [2] https://github.com/freebsd/freebsd-src/blob/7cd7b849b0629a259f4034514cf0e47e82efc44a/sys/dev/vt/vt_core.c [3] https://github.com/torvalds/linux/blob/cec1e6e5d1ab33403b809f79cd20d6aff124ccfe/drivers/tty/vt/vt_ioctl.c#L776 [4] https://github.com/illumos/illumos-gate/blob/92ae099e204069d2fec11f099863387b5317d849/usr/src/uts/common/io/vcons.c#L633
Seems reasonable. I'm curious how you came across this - did some code try (and fail) setting frsig?
(In reply to Ed Maste from comment #1) I am trying to get SDDM to start a Plasma Wayland session. I was stepping through the source and found that a call to VT_SETMODE fails, due to rtsig not being set and acqsig/relsig not being in range of NSIG. The latter is to be addressed in SDDM but omitting the frsig check here would help compatibility as there would be more programs coming from Linux that did not assume this quirk. [1] https://github.com/sddm/sddm/blob/fbffb577796c471e4fabf64f6422aa7bab4ea990/src/common/VirtualTerminal.cpp#L90
+1. I've bumped into this a few times while debugging standalone apps wanting to render without X11/Wayland e.g., https://github.com/mpv-player/mpv/pull/7647 https://github.com/vkmark/vkmark/pull/29
Interestingly, in lib/libvgl/main.c: smode.frsig = SIGINT; In any case I concur that frsig is currently unused and is unlikely to ever be used, so requiring userland to set it doesn't make much sense. I don't like having possibly-uninitialized values passed in though; what about frsig == 0 as a special case, e.g.: if (!(ISSIGVALID(mode->relsig) && ISSIGVALID(mode->acqsig) && (mode->frsig == 0 || ISSIGVALID(mode->frsig)))) { DPRINTF(5, "error EINVAL\n"); return (EINVAL); }
frsig == 0 wouldn't help portability e.g., mpv didn't set frsig at all, vkmark wanted acqsig == 0 and relsig == 0 as well. According to sys/compat/linux/linux_ioctl.c invalid signals are reset to 0 instead of returning EINVAL.
It looks like Linux doesn't care, it just passes whatever was provided in acqsig or relsig to kill_pid(). For us the Linuxulator behaviour makes sense because it has to translate the requested Linux signal to FreeBSD and there is no mapping for an invalid signal. mpv currently has `struct vt_mode vt_mode = { 0 };` and I'd expect that sort of idiom to be pretty common. sddm has vt_mode setModeRequest { }; bool ok = true; setModeRequest.mode = VT_PROCESS; setModeRequest.relsig = RELEASE_DISPLAY_SIGNAL; setModeRequest.acqsig = ACQUIRE_DISPLAY_SIGNAL; if (ioctl(fd, VT_SETMODE, &setModeRequest) < 0) { qDebug() << "Failed to manage VT manually:" << strerror(errno); ok = false; }
Created attachment 264332 [details] Proposed change ^Triage: rebase patch.
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=5198c32210039d8dc92554647384eee75688848c commit 5198c32210039d8dc92554647384eee75688848c Author: Ed Maste <emaste@FreeBSD.org> AuthorDate: 2025-09-30 19:39:34 +0000 Commit: Ed Maste <emaste@FreeBSD.org> CommitDate: 2025-11-27 00:19:04 +0000 vt: Allow VT_SETMODE with frsig=0 Linux does not check that any of the signals in vt_mode VT_SETMODE ioctl (relsig, acqsig, frsig) are valid, but FreeBSD required that all three are valid. frsig is unusued in both Linux and FreeBSD, and software typically leaves it unset. To improve portability, allow frsig to be set to zero. PR: 289812 Reported by: Dušan Gvozdenović Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D52835 sys/dev/vt/vt_core.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
A commit in branch stable/15 references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=224d65015465d085f2e07edccef1f23a8c217b88 commit 224d65015465d085f2e07edccef1f23a8c217b88 Author: Ed Maste <emaste@FreeBSD.org> AuthorDate: 2025-09-30 19:39:34 +0000 Commit: Ed Maste <emaste@FreeBSD.org> CommitDate: 2025-12-09 16:32:10 +0000 vt: Allow VT_SETMODE with frsig=0 Linux does not check that any of the signals in vt_mode VT_SETMODE ioctl (relsig, acqsig, frsig) are valid, but FreeBSD required that all three are valid. frsig is unusued in both Linux and FreeBSD, and software typically leaves it unset. To improve portability, allow frsig to be set to zero. PR: 289812 Reported by: Dušan Gvozdenović Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D52835 (cherry picked from commit 5198c32210039d8dc92554647384eee75688848c) sys/dev/vt/vt_core.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
^Triage: committed via D52835.
A commit in branch stable/14 references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=e1c985ccc335beb37bb34b368c5c019cc5f80f9c commit e1c985ccc335beb37bb34b368c5c019cc5f80f9c Author: Ed Maste <emaste@FreeBSD.org> AuthorDate: 2025-09-30 19:39:34 +0000 Commit: Ed Maste <emaste@FreeBSD.org> CommitDate: 2025-12-12 18:13:13 +0000 vt: Allow VT_SETMODE with frsig=0 Linux does not check that any of the signals in vt_mode VT_SETMODE ioctl (relsig, acqsig, frsig) are valid, but FreeBSD required that all three are valid. frsig is unusued in both Linux and FreeBSD, and software typically leaves it unset. To improve portability, allow frsig to be set to zero. PR: 289812 Reported by: Dušan Gvozdenović Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D52835 (cherry picked from commit 5198c32210039d8dc92554647384eee75688848c) (cherry picked from commit 224d65015465d085f2e07edccef1f23a8c217b88) sys/dev/vt/vt_core.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)