BSD Commit Log Search

   Update clzdi2.c to pull in the following commit neede because of a change
   in clang brought in with the llvm 22 update:

   commit 5d0e26e571c08dc4c0b2a25ed6c9f845f054fa76
   Author: Koakuma <koachan at protonmail.com>
   Date:   Tue Apr 29 07:36:32 2025 +0700

       [compiler-rt] Make sure __clzdi2 doesn't call itself recursively on sparc64 (#136737)

       On 64-bit platforms, libgcc doesn't ship with __clzsi2, so __builtin_clz
       gets lowered to __clzdi2. A check already exists for GCC, but as of
       commit 8210ca019839fc5430b3a95d7caf5c829df3232a clang also lowers
       __builtin_clz to __clzdi2 on sparc64.

       Update the check so that building __clzdi2 with clang/sparc64 also
       works.

   ok tb@, deraadt@

   Unbreak and update powerpc retguard for llvm 22

   For RETGUARD_LOAD_COOKIE when -fno-pie (as in macppc kernels), change
   a relocation in -fno-pie (as in macppc kernels) from unusual S_HIGHA
   "__retguard_3671 at higha" to usual S_HA "__retguard_3671 at ha".  This
   prevents an error from lld 22,

   ld: error: rasops15.o:(function rasops15_init: .text+0x2): unknown \
   relocation (111) against symbol __retguard_3671

   For RETGUARD_LOAD_PC in PIC code, change an instruction from 'bl .+4'
   to 'bcl 20,31,.+4' to fix branch prediction.  This follows the same
   change in upstream llvm,
   https://github.com/llvm/llvm-project/issues/128644

   ok jca@ naddy@

   And now fix the uvm page fault of OF_getpropstr() the right way.

   Correct offsets for redrawing entire line here too.

   Fix uvm fault panic introduced by new OF_getpropstr() function.

   Reported by jmc@.

   ok kettenis@

   bettertls/verify: resuls -> results

   sync

   revert upstream commit 947b9f55b5f327e14368a48fb6ce10242ea29bf3

   It breaks cpp -traditional -dM

   https://github.com/llvm/llvm-project/issues/200085

   add SipHash.h from llvm-22.1.6 which is shipped as a third party header

   sync after updating llvm to 22.1.6

   update build infrastructure for llvm-21.1.6

   thanks for all the help with testing and fixing things up

   merge lldb-21.1.6

   import lldb from LLVM-21.1.6

   merge lld-21.1.6

   import lld from LLVM-21.1.6

   merge clang-21.1.6

   import clang from LLVM-21.1.6

   merge llvm-21.1.6

   import llvm from LLVM-21.1.6

   Make qwx show the firmware's regulatory domain in debug logs, like iwx does.

   Add support for 40MHz channels to qwx.

   Also fix setting of 11n phy parameters during the association flow.
   The NODE_HT flag isn't set yet when qwx_assoc() runs. We need to apply
   11n-related peer parameters in qwx_run().

   Tested by phessler@ and myself.

   ok mlarkin@ phessler@

   Make qwx only allocate a new mbuf when a free Rx ring slot is available.

   Avoids pointless alloc and free when the ring is already full.

   avoid calling bus_dmamap_create() from interrupt context in qwx

   ok mlarkin@ phessler@

   Handle country code events sent by qwx firmware.

   Seems to fix the issues where association was unreliable.

   Tested by phessler@, robert@, and myself.

   ok mlarkin@ phessler@

   Use correct X position for redrawing wide characters.

   Note that the workaround for the ambiguous return value can also be used for timegm()

   OK tb@

   free exec_package allocations in check_exec() error paths

   from Andrew Griffiths at Calif, ok kettenis@

   acme-client: use _get0_ instead of _getm_ for const correctness

   The revokeproc doesn't modify the Validity, so use a const ASN1_TIME and
   switch to X509_get0_notBefore()/X509_get0_notAfter().

   suggested by job a long time ago

   acme-client: fix timegm(3) invocations

   Use the tm.tm_wday = -1 sentinel check and switch X509notbefore() and
   X509notafter() to use a time_t * output parameter since an in-band error
   doesn't work for them for the same reason it doesn't work for timegm(3).
   The error check remains against -1 to keep the diff small, but should
   perhaps be changed to use a Boolean 0/1 approach.

   Of course, with the aggressive push to very short-lived certificates in
   the WebPKI, negative epoch times won't happen here in practice.

   ok florian

   vmd: Unbreak using images supplied with vmctl -b

   My previous fix for reused filedescriptors broke using images
   supplied with "vmctl -b": In that case, kernfd remained to be set
   to -1, thus passing that filedescriptor to the child process failed.

   To resolve this, dup vm->vm_kernel right before passing the descriptor
   with proc_compose().  This fixes -b and the keeps original fix
   working.

   In the error path I rely on vm_stop()/vm_remove() closing vm->vm_kernel.
   Therefore, remove the redundant close().

   ok mlarkin@ deraadt@