Use lfence to serialize rdtsc for VIA Nano and up, same as Linux do.
Unlike Linux I didn't add a 64-bit guard on assumption that it should work
the same way and the 64-bit version will likely be the preferred one by users.
planning to pull up to netbsd-11, netbsd-10.
PR kern/68010 - sys/conf/osrelease.sh - add comments
Add some comments to osrelease.sh to explain its purpose,
keep away gawkers, etc...
While here, fix a couple of sh quoting botches.
NFCI.
No pullups.
Include model 7 in CPUID_TO_MODEL() extended model check.
Accommodates Zhaoxin CPUs using extended model.
Update the comment accordingly and fix one typo in it.
evbarm/INTEGRATOR_CP: establish interrupt handler for PL181 controller
PR/60818 plmmc_ifpga.c does not establish interrupt handler for PL181 controller
Capture the handle returned from ifpga_intr_establish - someone might
write the code to call ifpga_intr_disestablish.
fssconfig(8): some minor improvements
Try to clarify snapstore a bit more with more explicit language.
The proverbial desperate sysadmin at 3am doesn't need an exegetic
exercise.
fssconfig(8): some minor improvements
Avoid understcore in fss_dev, as it doesn't mesh well with the
underlining used for .Ar in plain text.
Make [-c] case visually match the other "action" options -l, and -u.
risc-v: remove initialisation of ci_shootdowncpus as it's unused.
RISC-V (currently) uses SBI services to invalidate TLB entries so
ci_shootdowncpus is unused. Remove its initialisation, but retain
the field to avoid changing struct cpuinfo layout, and it may be
used in the future.
Handle sync, interlace in GTF modes, explicity init name in EDID and GTF.
Avoid stack garbage passed in those uninitialized fields - fixes
wsdisplay drivers that actually read these fields.
Fixes and features for PPC440 core, PPC460EX chip, Sam460ex board.
- Add ebc - PPC440 family External Bus Controller driver.
- Add samfpga - Driver for Sam460ex on-board FPGA (provides 80 GPIOs,
soft power off, 3x led(4), etc. Lives on the ebc.
- Enable PPC460EX on-chip GPIO via gpio_opb.
- Implement a bunch of workarounds for 440/460 hardware bugs.
- Enable pm2fb in SAM460EX config (necessary fixes to pm2fb pending
commit).
- Disable OHCI in SAM460EX config - due to hardware bug operating
two AHB masters at the same time may cause lockups. And it sure does,
there's no known software workaround.
- Don't allow passing PSL_FP from MSR userland bits (userland can cause
panic).
- Fix ghost interrupt deliveries where asserted line re-entered the
exception.
- Ensure interrupt level-mask flag is set by default.
- Tag mmaped DMA address as cacheable unless memory was explicitly
allocated as uncached (only for 440, OEA probably could use it too).
[11 lines not shown]
Add Centaur cpuid and family 7 to check whether the TSC counter is invariant.
Recent Zhaoxin CPUs have an invariant TSC and advertise that through
CPUID 80000007, as Intel and AMD do.
Allows choosing the TSC timecounter as more reliable than HPET.
XXX pullup-11, and perhaps netbsd-10.
mips: create ci_shootdowncpus early so it's available for MP pmap_update
When using ENABLE_MIPS_4KB_PAGE on OCTEON with an MULTIPROCESSOR kernel
an early pmap_update is triggered which requires ci_shootdowncpus to be
vailable.
A bunch of fixes to wscons accumulated over last few months.
Glyph cache:
- Handle situation when it can not hold a single bucket -
fixes garbage after loading a font that does not fit in VRAM.
- Make sure that glpyhcache_wipe() clears the whole map -
fixes wrong glyphs after a font change when cache is small.
- Return GC_NOPE when cache has no buckets - fixes NULL deref
when someone tries to draw before initializing the cache.
- Don't allow glyphcache_try() to steal when only one bucket
exists - fixes corruption for second color on small caches.
vcons:
- When loading font only call show_screen_cb for visible
screen - fixes corrpution of visible screen when loading font
for another screen.
- Make underlined blanks go through putchar() instead of
erasecols() - fixes loss of underlined spaces on screen
redraw.
[12 lines not shown]
mips: remove the OCTEON workaround for PR/59236 as fixed by PR/60664
The rdhwr emulation bug observed in PR/59236 was fixed by the patch in
PR/60664 so the workaround applied for PR/59236 can be safely removed.
