BSD Commit Log Search

   Fix the mcount entry name in the 68010 case.

   nix whitespace

   Welcome to 11.1!

   Disable 11.0 dedication

   Decorate next68k_bootargs() with no_instrument_function.

   new expat 2.10.0 out

   Add -fno-unwind-tables to CFLAGS on all m68k platforms.

   m68k/conf/Makefile.kern.m68k.inc now sets -fno-omit-frame-pointer
   automagically when it's needed, so don't bother doing it in
   individual kernel configs.

   Also set -fno-omit-frame-pointer if DDB or KGDB is configured.

   just avoid hardcoding the interface key.

   use the name variable and adjust for it in kernel.

   needs both userland and kernel update to be in sync

   revert last diff

   passed wrong nvlist key to the kernel

   fixes broken tests

   Profiling requires the frame pointer, so if PROF is set in the Makefile,
   add -fno-omit-frame-pointer.

   Problem reported by Romain Dolbeau on port-m68k.

   i915: Fix execlist submission register indexing harder.

   This time, with any luck, I transcribed the indexing correctly.

   PR port-amd64/58657: i915drmkms blank screen

   atto(1): Errors relating to the menu generally aren't fatal,
   so don't treat them as such.

   atto(1): fatal cannot take special format string arguments other than
   the initial %s for the program name. Gain some safety by not having
   callers pass a format string in the first place.

   With this, we can raise WARNS to 5.

   atto(1): don't cast malloc return value

   Revert the reclaimd changes until I've had a chance to look into the
   ramifications more closely.

   PR lib/60949 pthread*_getname_np(3) - avoid UB if len == 0

   Just make len == 0 be an EINVAL error.

   Update pthread_attr_getname_np(3) man page, so it no longer says
   there are no errors from pthread_attr_getname_np() - which was wrong
   in any case, as there was already one (if attr is invalid), and
   while there, one minor grammar fix.

   No pullups currently planned.

   Correct mapping of KS_Power for the "Keyboard Power" Usage ID.

   It's USB Keyboard/Keypad Page usage 102 (not 109 which is F18).

   In the fpfline trap handler, use the faultstkadj path rather than the
   basic fault path.  This is to handle FPU emulation faults on the 68LC040,
   which use a format 4 frame; if the emulated FPU insn results in a signal,
   kernel stack would be leaked.

   Problem reported by Romain Dolbeau on port-m68k.

   Allocate state for the FPU emulator on the kernel stack rather than as
   static globals.  Normally we wouldn't want to allocate this much on the
   stack, but the emulator is only for user-space code, so the kernel stack
   will have been idle when the F-line trap occurred; the only things on
   the stack at this point are the trap frame and the small number of call
   frames required to get us here.

   This prevents concurrent FPU emuations from colliding with each other.
   This is something that could happen when e.g. two FP-using programs are
   running concurrently on a system under memory pressure and one of them
   blocks during an emulation to service a page fault.

   The alternatives are less than ideal: we either have to dynamically
   allocate memory or wrap the emulator in a mutex.

   Similarly, change __fpu_logn() and __fpu_modrem() to place their
   temporaries on the stack, as well.

   Problem reported by Romain Dolbeau on port-m68k.

   Don't proceed with undefined results in the face of memory access
   errors in the FPU emulator:
   - fpu_load_ea() / fpu_store_ea(): if copyin() / copyout() return non-zero,
     return SIGSEGV.

   Catch some callers of fpu_load_ea() and fpu_store_ea() that were ignoring
   the return value:
   - fpu_emul_arith() / fpu_emul_fscale(): Pay attention to the return value
     of fpu_load_ea().
   - fpu_emul_fstore(): Pay attention to the return value of fpu_store_ea().

   Problem reported by Romain Dolbeau on port-m68k.

   i915: Fix execlist submission register indexing.

   PR port-amd64/58657: i915drmkms blank screen

   secmodel.9: update some details in code examples

   Back in 2009, refactoring changed where certain functionality is
   located, but the examples in this man page weren't updated to match.

   prevent invalid sequences being written out, instead print '?' (Kilian Feess)

   prevent negative offsets hiding bugs (Kilian Feess)

   widen types to prevent truncation (Kilian Feess)

   Add checks to unchecked allocations (Kilian Feess)

   avoid buffer overreads (Kilian Feess)

   set nscrap to 0 on malloc failure (Kilian Feess)