From Eric Huang
51ccaf0e30c303149244c34820def83d74c86288 in linux-6.18.y/6.18.21
14b81abe7bdc25f8097906fc2f91276ffedb2d26 in mainline linux
This requires a further change to avoid a sleeping alloc while
holding a mutex. Not yet in 6.18.y
drm/amdgpu: fix the idr allocation flags
From Prike Liang
62f553d60a801384336f5867967c26ddf3b17038 in mainline linux
This fixes state machine issues that resulted in hangs with OpenBSD-i386 VMs
when using the i8254 hardware timecounter with vmm. This also manifested in
incorrect i8254 calibration (wrong CPU speed in dmesg).
tested with Debian 12 amd64, i386
Apine Linux 3.23 x86 (with 4GB memory), x86_64
OpenBSD 7.9 beta amd64, i386
ok mlarkin@
Zero the DM PTE/PDE pages before use. Fixes a bug on machines with more than
512GB RAM; those pages might contain previous data/junk and panic during
pmap_randomize.
Tested on various amd64 laptops, an openbsd amd64 vmm VM and an EPYC server
with 1TB RAM.
Fix supplied by Chris Cunningham, thanks!
Fix a bug in relay_inherit() which runs only
relay_load_certfiles(conf, rb, NULL) unconditionally which isn't
alligned with logic in parser when it parses relay block, where multiple
certificates are load as relay_load_certfiles(conf, rb, NULL) only if
here no tlscerts (for default host) and otherwise it loads keypairs.
OK: rsadowski@
x509v3.h has included conf.h since June 20, 1999, OpenSSL commit ba404b5e,
so HEADER_CONF_H has been defined since then. Also since then, CONF_VALUE
(only available via conf.h) has been used outside of HEADER_CONF_H, making
that #ifdef doubly pointless.
ok bcook jsing kenjiro
All thirteen files including cms_local.h do that after including cms.h,
which already includes x509v3.h, so this is always defined. While here
make the cms_local.h a bit more selfstanding by including asn1.h and
x509v3.h
ok bcook jsing (who had the same diff) kenjiro
special register containing its value, this is a general register (x4)
and therefore can be used directly
Diff stolen from miod@ who is on strike but agreed to let me commit
this. ok kettenis@
If the tag entry points to the line number zero, less crashes because
internally it uses that line number to imply that the tag is associated
with a pattern.
issue reported by Henry Ford (henryfordkjv at gmail), thanks!
ok kirill@
extended data, because that may still be in use. Add a flag to
grid_clear_cell to indicate this. Fixes irritating problems with ICH
(CSI @) mostly visible in emacs.
Increment ncpus (hw.ncpu) in mips64 cpuattach() instead of octeon/loongson
hw_cpu_hatch() running before, not after MI sofnet_percpu(), respectively.
This matches what arm64, macppc and powerpc64 do.
Local traffic may now be distributed across multiple threads, while physical
drivers like cnmac(4/octeon) still lack multiqueue support to do so.
loongson was not tested, but should behave the same.
initial report Jordan Geoghegan
tests kirill bluhm Janne Johansson
feedback miod bluhm kettenis
OK kirill visa
flushing code to operate on virtual addresses instead of physical
addresses. Seems the Zicbom implementation on the SpacemiT X60 cores
doesn't flush the caches if the mapping is non-cachable. So adjust
_pmap_kenter_pa() to use a temporary cachable mapping to clean a page we
want to map non-cachable.
ok jca@
