BSD Commit Log Search

   Add some stage-2 defines

   Print some more EL2 system registers in DEBUG_LOCORE

   Merge the m68k trap code.  Still a bit more tidying up to do here, but
   this eliminates a significant chunk of code duplication.

   Biggest change here is for Amiga, whose different-from-everyone-else's
   68040 write-back code was voted off the island.

   Fix case of normal numbers starting with a highword of 0x0 for m68k.

   In the case of mac68k there can still be normal numbers with a highword
   value of 0.  This differs from x86 where they would be considered denormals.

   No need for special handling for m68ksf.

   Also the previous LDBL_MIN was a denormal on m68k...opps.

   libc: test long double conversions

   On sparc64, tests/usr.bin/xlint/lint1/msg_380.c and msg_381.c fail.

   libcurses: fix wrong NULL guard

   Move the NULL guard in _cursesi_waddbytes() out of #ifdef DEBUG .
   Delay initialization of all variables that require dereferencing pointers
   that might be NULL until after the NULL guards.  This is what KNF calls
   for anyway.

   Fixes PR lib/60153 .

   FAFO

   Tweak an pmaphist log message

   Use pmapxtabhist in this file as the functions are called A LOT

   arm: revert the other part of "arm: optimize splx"

   The ci_pending_ipls test is racy too.

   arm: revert part of "arm: optimize splx"

   Testing ci_intr_depth to detect recursion in splx is wrong.

   gicv3: Stop using ICC_PMR_EL1 to mask interrupts.

   Managing ICC_PMR_EL1 with spl changes is too costly, even with lazy PMR
   updates. Let's simplify things. We can still take advantage of interrupt
   prioritization for delivery but let's do spl filtering in software.

   On Ampere eMAG, iperf3 on the loopback interface sees a ~18% performance
   improvement with this change.

   arm: optimize splx

   Skip masking / unmasking interrupts and pic_do_pending_interrupts if
   there is another pic_do_pending_interrupts in progress or if there are
   no pending unmasked IPLs on the CPU.

   arm: Make __HAVE_PIC_PENDING_INTRS state per-CPU.

   Move pic_pending_irqs / pic_blocked_irqs / pic_pending_ipls from per-PIC
   to per-PIC-per-CPU state. This allows us to defer processor-local
   interrupts like SGIs and PPIs that use the same irq number.

   mi pmap: Rename two functions for clarity of purpose

   tlbinfo_asids_p -> tlbinfo_hasasids_p
   This functions indicates if ASIDs are implemented

   tlbinfo_noasids_p -> tlbinfo_freeasids_p
   The new function indicates if there are free ASIDS in the pool (the
   inversion of the previous function)

   fix few typos in comments.

   fix typo in the referenced filename (in comment).
   s/makstrerror.awk/makestrerror.awk/

   s/A_BCM1480_SWTRC_MATCH_TAG_MAKS_0/A_BCM1480_SWTRC_MATCH_TAG_MASK_0/ in macro
   constant definition.

   It is later used with correct name in A_BCM1480_SWTRC_MATCH_TAG_MASK macro.

   add hints about regenerating dtb set lists

   rdcpcib(4): fix rdcpcib attach/detach handling for function 0 and function 1.

   Only function 0 exposes the indirect access registers and the watchdog timer.
   Function 1 has a separate register set and does not use these registers.

   This resolves the autoconfiguration error seen in boot dmesg:
   "rdcpcib1: autoconfiguration error: couldn't map indirect registers"

   Attach now maps the indirect access registers and configures the watchdog only
   for function 0, and detach only releases resources that were initialized.

   reviewed by riastradh@, thanks!

   lint: on sparc64, investigate test failure in msg_380 and msg_381

   Fix tyop in comment.

   xz-5.8.3 out

   whitespace

   swread/swwrite: update a comment about swapdev stability

   https://gnats.netbsd.org/60147
   PR/60147

   Fill in bus_dmamap_sync() for 68030.  Since 68030 caches are write-through,
   this is pretty trivial: just invalidate the on-chip + any external D cache
   in the PREREAD case.  (See comment in code for why PREREAD is preferred
   over POSTREAD.)

   PR kern/60144

   uvm_swap.c: fix /dev/drum regression caused by swap encryption

   * decrypt data using a bounce buffer

   * reject writes with ENOTSUP for now

   https://gnats.netbsd.org/60083
   PR/60083

   swread/swwrite: hold swap_syscall_lock (cont.)

   fix a harmless merge botch i introduced while porting the patch
   from git to cvs.

   https://gnats.netbsd.org/60147
   PR/60147

   regenerate dtb lists