Date: Sun, 22 Mar 2026 20:01:04 +0200 From: Konstantin Belousov <kostikbel@gmail.com> To: A FreeBSD User <freebsd@walstatt-de.de> Cc: src-committers@freebsd.org, dev-commits-src-all@freebsd.org, dev-commits-src-main@freebsd.org Subject: Re: git: 8cc1c0f35ec8 - main - amd64 trap.c: provide tag for the struct sfhandlers definition Message-ID: <acAuYEscAN_FAQ4M@kib.kiev.ua> In-Reply-To: <20260322111330.6c1a3249@thor.sb211.local> References: <69bdcf15.27526.2b5005f0@gitrepo.freebsd.org> <20260321171438.680a77e3@hermann> <ab7qCc3Dx7TCnyAD@kib.kiev.ua> <20260321224023.002fcb9b@hermann> <ab8UjfCTLRBZpevm@kib.kiev.ua> <20260322111330.6c1a3249@thor.sb211.local>
index | next in thread | previous in thread | raw e-mail
On Sun, Mar 22, 2026 at 11:13:03AM +0100, A FreeBSD User wrote: > Am Tage des Herren Sat, 21 Mar 2026 23:58:37 +0200 > Konstantin Belousov <kostikbel@gmail.com> schrieb: > > > On Sat, Mar 21, 2026 at 10:40:23PM +0100, FreeBSD User wrote: > > > On Sat, 21 Mar 2026 20:57:13 +0200 > > > Konstantin Belousov <kostikbel@gmail.com> wrote: > > > > > > > On Sat, Mar 21, 2026 at 05:15:55PM +0100, FreeBSD User wrote: > > > > > On Fri, 20 Mar 2026 22:49:57 +0000 > > > > > Konstantin Belousov <kib@FreeBSD.org> wrote: > > > > > > > > > > > The branch main has been updated by kib: > > > > > > > > > > > > URL: > > > > > > https://cgit.FreeBSD.org/src/commit/?id=8cc1c0f35ec8d5a3edb4ee1ede962a3c1f6baef0 > > > > > > > > > > > > commit 8cc1c0f35ec8d5a3edb4ee1ede962a3c1f6baef0 > > > > > > Author: Konstantin Belousov <kib@FreeBSD.org> > > > > > > AuthorDate: 2026-03-03 06:37:03 +0000 > > > > > > Commit: Konstantin Belousov <kib@FreeBSD.org> > > > > > > CommitDate: 2026-03-20 22:49:03 +0000 > > > > > > > > > > > > amd64 trap.c: provide tag for the struct sfhandlers definition > > > > > > > > > > > > Sponsored by: The FreeBSD Foundation > > > > > > MFC after: 1 week > > > > > > --- > > > > > > sys/amd64/amd64/trap.c | 6 ++++-- > > > > > > 1 file changed, 4 insertions(+), 2 deletions(-) > > > > > > > > > > > > diff --git a/sys/amd64/amd64/trap.c b/sys/amd64/amd64/trap.c > > > > > > index 8a1d1528e6da..359b3dfe3609 100644 > > > > > > --- a/sys/amd64/amd64/trap.c > > > > > > +++ b/sys/amd64/amd64/trap.c > > > > > > @@ -319,10 +319,12 @@ trap_check_intr_kernel(struct thread *td, struct > > > > > > trapframe *frame) /* > > > > > > * Table of handlers for various segment load faults. > > > > > > */ > > > > > > -static const struct { > > > > > > +struct sfhandler { > > > > > > uintptr_t faddr; > > > > > > uintptr_t fhandler; > > > > > > -} sfhandlers[] = { > > > > > > +}; > > > > > > + > > > > > > +static const struct sfhandler sfhandlers[] = { > > > > > > { > > > > > > .faddr = (uintptr_t)ld_ds, > > > > > > .fhandler = (uintptr_t)ds_load_fault, > > > > > > > > > > > > > > > > Hello, > > > > > > > > > > it seems that one of the commits in this vain results in a hard reset of > > > > > some CURRENT machine - this taken from some reports on freebsd-current list. > > > > > > > > > > In my case, two Zen5 based boxes, bott as usual, but when touch the system > > > > > with "make buildworld buildkernel" or any single make build on the acting > > > > > server box, the system reboots immediately - no trace of a log, coredump, > > > > > simple nothing. Only hard reset. Same on another box acting as graphical > > > > > workstation. As soon the system gets "touch" (login/starting windowmaker) > > > > > the system goes nirwana. > > > > > > > > > > No further information, my boxes do not dump anything useful. > > > > > > > > Try the following > > > > > > > > commit 237044fd4d0659ef2799c259b1f1852c00b08b37 > > > > Author: Konstantin Belousov <kib@FreeBSD.org> > > > > Date: Sat Mar 21 20:45:48 2026 +0200 > > > > > > > > amd64: use IDT trapframe size to calculate location of nmi_pcpu from the > > > > IST stack pointer > > > > > > > > diff --git a/sys/amd64/amd64/exception.S b/sys/amd64/amd64/exception.S > > > > index babfbacf92ef..5bf9a12e1bf5 100644 > > > > --- a/sys/amd64/amd64/exception.S > > > > +++ b/sys/amd64/amd64/exception.S > > > > @@ -341,7 +341,7 @@ IDTVEC(dblfault) > > > > pushfq > > > > andq $~(PSL_D | PSL_AC),(%rsp) > > > > popfq > > > > - movq TF_SIZE(%rsp),%rdx > > > > + movq TF_SIZE_HW(%rsp),%rdx > > > > movl %edx,%eax > > > > shrq $32,%rdx > > > > movl $MSR_GSBASE,%ecx > > > > @@ -661,7 +661,7 @@ IDTVEC(dbg) > > > > shlq $32,%rdx > > > > orq %rdx,%r12 > > > > /* Retrieve and load the canonical value for GS.base. */ > > > > - movq TF_SIZE(%rsp),%rdx > > > > + movq TF_SIZE_HW(%rsp),%rdx > > > > movl %edx,%eax > > > > shrq $32,%rdx > > > > wrmsr > > > > @@ -798,7 +798,7 @@ IDTVEC(nmi) > > > > shlq $32,%rdx > > > > orq %rdx,%r12 > > > > /* Retrieve and load the canonical value for GS.base. */ > > > > - movq TF_SIZE(%rsp),%rdx > > > > + movq TF_SIZE_HW(%rsp),%rdx > > > > movl %edx,%eax > > > > shrq $32,%rdx > > > > wrmsr > > > > @@ -1002,7 +1002,7 @@ IDTVEC(mchk) > > > > shlq $32,%rdx > > > > orq %rdx,%r12 > > > > /* Retrieve and load the canonical value for GS.base. */ > > > > - movq TF_SIZE(%rsp),%rdx > > > > + movq TF_SIZE_HW(%rsp),%rdx > > > > movl %edx,%eax > > > > shrq $32,%rdx > > > > wrmsr > > > > diff --git a/sys/amd64/amd64/genassym.c b/sys/amd64/amd64/genassym.c > > > > index 2716784ee871..762c850c1df1 100644 > > > > --- a/sys/amd64/amd64/genassym.c > > > > +++ b/sys/amd64/amd64/genassym.c > > > > @@ -185,6 +185,7 @@ ASSYM(TF_FS, offsetof(struct trapframe, tf_fs)); > > > > ASSYM(TF_GS, offsetof(struct trapframe, tf_gs)); > > > > ASSYM(TF_FLAGS, offsetof(struct trapframe, tf_flags)); > > > > ASSYM(TF_SIZE, sizeof(struct trapframe)); > > > > +ASSYM(TF_SIZE_HW, sizeof(struct trapframe) - 2 * 8); > > > > ASSYM(TF_HASSEGS, TF_HASSEGS); > > > > > > > > ASSYM(PTI_RDX, offsetof(struct pti_frame, pti_rdx)); > > > > > > > > > > Just gave the patch of yours a chance before sleeping time on the GUI box of > > > mine: after reboot, starting GUI (windowmaker) a total freeze of the system > > > immediately. > > > I'll check on the other machine without burden of GUI tomorrow morning. > > > > > > Still it seems the box isn't dumping/reporting anything, this time just frozen. > > > > Instead of the patch I posted above, try the following: > > diff --git a/sys/x86/include/frame.h b/sys/x86/include/frame.h > > index 7da1fff03aec..1560983addfa 100644 > > --- a/sys/x86/include/frame.h > > +++ b/sys/x86/include/frame.h > > @@ -153,8 +153,10 @@ struct trapframe { > > uint16_t tf_fred_evinfo1; > > uint32_t tf_fred_evinfo2; > > /* two long words added by FRED */ > > +#if 0 > > uint64_t tf_fred_evdata; > > uint64_t tf_fred_zero1; > > +#endif > > }; > > > > #define TF_FRED_EVDATA_B0 0x0000000000000001ull /* %dr6 B0 */ > > > > BTW, did you rebuild the drm kmod with any of the kernel updates? > > > > Good things first: latest sources with your commit 1ba29614c4ce5e261ade0bd7def94079b7b9647a > fixes my issues. > > Just for the record: applied the patch on the other box as, it freezes > immediately when trying to login or, in single user, when confirming the shell at the > beginning. > Then there is some issue with the metodology of the testing. What I committed is the equivalent of the patch above, and according to your report, and other people mails, it fixed the issue, as I expected. > And: yes, I try to build on every buildworld/buildkernel cycle ports essential to kernel > dependencies (especially kernel modules like GPU driver) I'm aware of. > > > Thank you for taking your time and fixing the problem! > > Best regards > Oliver > > > > > -- > > A FreeBSD userhome | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?acAuYEscAN_FAQ4M>