Update PMAP_STEAL_MEMORY code to uvm_hotplug

(skrll)

Add '_p' to uvm_physseg_valid to match the function.  Bump date.

(skrll)

Print nvm version of ixl(4) when attaching

reviewed by msaitoh

(yamaguchi)

Use flags instead of bool that changes behavior related to
adminq command of ixl(4)

(yamaguchi)

Fix the wrong expression in ixl(4)

Closes PR/54860.

(yamaguchi)

support internal PHY temperature sensor

(ryo)

Remove npf in POGO because it's now in SHEEVAPLUG.

(jklos)

make counters per queue

(ryo)

set/clear IFF_OACTIVE flag only on txring 0

(ryo)

Add Fast Short Rep Mov(fsrm).

(msaitoh)

Fix a bug that the virtual channel extended configuraton's arbitration phase
register can't be decoded correcty. Found by jmcneill.

(msaitoh)

Don't remove the test harness utilities (the tests were not removed already
which is the majority of the files), we can use them for atf now.

(christos)

Add back the test harness

(christos)

make things compile (pullup merged)

(christos)

fix test I broke by commenting out flags.

(christos)

pull fix from head (O_EXEC means O_SEARCH only on FreeBSD)

(christos)

Fix /tmp tmpfs handling and on machines with enough RAM default to
creating a /tmp tmpfs with 25% of ram size limit. Suggested by ad@.

(martin)

Push the vnode locking in namei() about as far back as it will go.

(ad)

Initialize the session variable to an impossible session to prevent compiler
warnings.

(christos)

Remove useless calls to get_ramsize() - return value is ignored and
the function has no side effects.

(martin)

Use .org directive to fix disklabel location instead of hand calculation.
If the code size before disklabel is less, zero is filled automatically.
If the code size before disklabel is exceeded, assembler stops with error.
Note that this changes a padding word in the current output binary from
a NOP instruction to zero but there are no functional changes.

(isaki)

use the CARP interface for arp/nd instead of the carp parent interface.
this provides the correct source mac address for the packets.

there are routers out there that cache the source mac during
nd and then subsequently bypass/miss packet filters on carp
interfaces as they send to the parent interface mac instead of the
correct carp interface mac.

(kardel)

Rewrite limit check using cross-build tools.
This eliminates use of hosts tr(1) and bc(1).
(But I'm going to remove this again soon due to other reason)

(isaki)

Provide SIOCGIFMEDIA ioctl to deliver link status.
Add link0 (IFF_LINK0) flag to map INIT state to LINK_STATE_DOWN
instead of LINK_STATE_UNKNOWN. This allows routing software to
suppress routes to the interface of the carp interface when in
init state (e. g. link down in the parent interface).

(kardel)

Remove .else part of .if OBJECT_FMT == ELF.
All similar parts in other Makefiles are removed in 2010.

(isaki)

Back out previous, it's broken.

(ad)

Fix wrong usage of PCI_PRODUCT(), ixl(4)

(yamaguchi)

dm(8) doesn't cause problems on suspend, don't block it.

Tested by Matthias Petermann, thanks!

(maya)

Fix ixl(4) not to assign interrupts to cpu #0
if ncpu > queue pairs

(yamaguchi)

Use def{flag,param} for parameters in ixl(4)

(yamaguchi)

make the topology support empty on non-SMP enabled kernels.
reduces kernel sizes and also fixes the builds.

(mrg)

Avoid using the global "pm" variable in utility functions - during
extended partitioning they may point to a different device.

(martin)

Add a method to query the partitioning schemes "internal idea" of a
cylinder size - whatever that means in the real world.

(martin)

Fix handling of inner/outer partitions (e.g. MBR and disklabel) in
extended partitioning:
- when editing a disk with such a schme, offer both partitions in two
  steps
- when commiting changes save the inner partitions after the outer ones

(martin)

Back out the cpu_switchto() workaround now that it's no longer needed.

(ad)

NetBSD 9.99.38 - VM system changes

(ad)

vmstat -s: report new stats:

        0 per-cpu stats one synced
    7246 per-cpu stats all synced
    4092 anon pages possibly dirty
    8881 anon pages dirty
        0 anon pages clean
      68 file pages possibly dirty
        0 file pages dirty
  2367889 file pages clean

(ad)

Merge from yamt-pagecache (after much testing):

- Reduce unnecessary page scan in putpages esp. when an object has a ton of
  pages cached but only a few of them are dirty.

- Reduce the number of pmap operations by tracking page dirtiness more
  precisely in uvm layer.

(ad)

fstyp: Remove redundant best_i check in HAMMER2

https://reviews.freebsd.org/D23159
taken-from: DragonFlyBSD originally from FreeBSD

(tkusumi)

fstyp: Use strlcpy(3) for HAMMER1

https://reviews.freebsd.org/D23159
taken-from: DragonFlyBSD originally from FreeBSD

(tkusumi)

Push the INVLPG limit for shootdowns up to 16 (for UBC).

(ad)

Add imxi2c

(jmcneill)

Add i2c clocks

(jmcneill)

Match the more generic fsl,imx21-i2c compat string

(jmcneill)

Rewrite bcmp() & memcmp() to not use REP CMPS.  Seems about 5-10x faster for
small strings on modern hardware.

(ad)

Fix build

(skrll)

add a note about arm* cpu topology support.

(mrg)

port the arm64 cpu topology setup for big.little to arm.

rename arm64 cpu_do_topology() to arm_cpu_do_topology() and
call it from both arm cpu_attach().

replace both aarch64_set_topology() inline code in arm
cpu_attach() with new arm_cpu_do_topology(), which is called
by the arm64 locore as well (possibly not needed, which would
allow it to become static.)

not yet tested on a real big.little armv7 system.  tested
on rockpro64 and pinebook pro.

(mrg)

Fix typo (s/txr_opackets/txr_obytes/)

pointed out by ryo@n.o, thanks.

(yamaguchi)

If malloc fails in zyd_node_alloc then return NULL.  From nick-nhusb

(skrll)

Use modern interfaces and kill tsleep/wakeup.  From nick-nhusb branch.

(skrll)

Trailing whitespace

(skrll)

Use a define instead of a magic constant.  NFCI.

(skrll)

tsleep -> kpause

(thorpej)

tsleep -> kpause

(thorpej)

evbarm: Add support for NXP i.MX 8M family SoCs.

(jmcneill)

Add support for NXP i.MX 8M Dual/8M QuadLite/8M Quad family SoCs.

(jmcneill)

.include bsd.own.mk for ${NETBSDSRCDIR}

(christos)

Add AMDFCH41_SMBUS_PORTMASK.

(msaitoh)

enable the BAD_NCQ quirk for all 860 EVO drives

XXX work-in-progress, it's not clear whether this is driver or controller
XXX problem

(jdolecek)

In the extract atomic case, remove the destination hard link first
since it did not get removed before for us. If we wanted to do this
right, we would do the mk*temp() file generation and trying to link
instead of open, and finally when we got the link to succeed, we
would rename the temp link to the final destination name, making
hard link creation atomic. For now we just remove first.

(christos)

PR install/54787: when trying to derive proper alignement and first offset
from exisiting partitions, use the first partition offset if it is already
closer to the start of the disk than the prefered alignment - we can not
move existing partitions around.

(martin)

Adjust the error return value of pthread_sigmask for !libpthread usage

Instead of returning -1, return errno on error.

Catch up after the fix in libpthread by Andrew Doran in 2008
in lib/libpthread/pthread_misc.c r.1.9.

It's an open question whether this function shall be used without linked
in the POSIX thread library.

Detected by Bruno Haible (GNU) and documented in gnulib in commit
"pthread_sigmask: Avoid test failure on NetBSD 8.0. " r. 4d16a83b0c1fcb6c.

(kamil)

Back to pax as tar until I figure out how I broke libarchive (hard link
extraction creates 0 length files for existing hard-links).

(christos)

add noinline attribute

(scole)

Update for AMD FCHs (HUDSON, BOLTON and KERNCZ).

(msaitoh)

Save/restore port number before selecting port. Linux driver says it must
be required on some systems.

(msaitoh)

Simplify. No functional change.

(msaitoh)

Catch up in ktruss(1) with ptrace descriptive operation names

Switch to dynamic string lists PT_STRINGS and PT_MACHDEP_STRINGS.

Reuse the code and approach from kdump(1).

(kamil)

Regen.

(ad)

namecache:

This is working better than expected.  It seems to cut system time for
build.sh by ~10% on my test machine and joerg@ is seeing better results with
pbulk.  Improve it a bit more without changing the basic idea:

- Split cache_list_lock into a per-vnode rwlock for reverse lookup, and a
  lightly contended global lock on LRU state (cache_lru_lock),

- For LRU replacement, imitate the VM system's page replacement algorithm.
  This eliminates the writebacks to struct namecache (to track time of last
  hit).

- Dynamically allocate the per-directory lock, preparing the way for having
  a "struct nchdir" or similar which could contain stuff like different
  structures for lookup, cached info to do the equivalent of VOP_ACCESS() in
  cache, and so on.

(ad)

No functional change:

- Move some definitions from if_stgereg.h to if_stge.c again because those are
  not chip (registers or descriptors) definitions.
- Use proplib to pass information that loading DSP code is required when
  PHY reset.

(msaitoh)

New sentence, new line.

(wiz)

Clean up.
- B_KEYINP is no longer called from C.
- B_COLOR is no longer used.

(isaki)

consio_init() should be done before printf().

(isaki)

If "application processors" were skipped/disabled at boot time (due to
RB_MD1 being set), don't try to examine the featurebus info, since it
was never retrieved.  Addresses kern/54815

XXX pullup-9

(pgoyette)

bump date for previous change

(jdolecek)

document the wd(4) sysctl nodes, and add the note about the Sumsung EVO drives

part of fix for PR kern/54790 and PR kern/54855

(jdolecek)

disable NCQ by default for "Samsung SSD 860 EVO 1TB" and
"Samsung SSD 860 EVO 500GB" - these drives have known broken NCQ support
particularly when used with AMD SB710/750 chipsets, problem occur also
under Linux and Windows

https://eu.community.samsung.com/t5/Cameras-IT-Everything-Else/860-EVO-250GB-causing-freezes-on-AMD-system/td-p/575813
https://bugzilla.kernel.org/show_bug.cgi?id=201693

It seems there is no Samsung firmware update to fix this even.

Disable NCQ regardless of the controller, it's likely same problem
exists with other controllers too.

This should fix PR kern/54790 and PR kern/54855

(jdolecek)

Revert previous; we don't want arm to have variable page size. Instead
nick@ is fixing the cache pages issue to use a separate pool.

(christos)

Fix some more bugs in the topo stuff, that prevented it from working
properly with fake topo info + MP.

(ad)

Don't call cpu_switchto() before idle_loop(), it should not be needed any more.
While there, assume (and KASSERT) that curlwp == ci->ci_data.cpu_idlelwp,
this saves a lwp_getpcb() call.

(bouyer)

Fix atfork malloc handling to include all the locks in old jemalloc.

(joerg)

Fix aarch64 definition in old jemalloc

There is no platform-specific reason for avoiding TLS. The tiny
allocations should be aligned the same as small allocations for ABI
reasons.

(joerg)

use the host compiler

(christos)

Rip out some very ambitious optimisations around pthread_mutex that are
don't buy much.  This stuff is hard enough to get right in the kernel let
alone userspace, and I don't trust that it's right.

(ad)

Fix mis-incrementing pointer size in bus_space_read_region_{4,8}

pointed out by jmcneill@. thanks.

(ryo)

Sort the list.  Minor markup and wording tweaks.

(uwe)

Oops.  Restore the dot lost in -width ".Dv ..."

(uwe)

Fix -width in the tagged list of special "handle" values.

(uwe)

Minor tweaks.

Use same -indent for lists of values that are OR'ed as one argument.
Use consistent markup for dli_* member names.
Untabify Dl_info definition so that it lines up in PS output.

(uwe)

Fix a typo

(kamil)

Correct the authorship note of dlinfo(3)

(kamil)

Miscellaneous fixes.

No space after asterisk in "type *arg".
There's no "-width indent", use something relevant instead.
Do not document the structure tag of the Link_map typedef.
Give the Link_map members a paragraph of their own for readability.
Give them a bit more breathing space in the tagged list.
Spell pooka's name correctly.

TODO: The text is pleonastic in several places ("Do foo.  On success
does foo." etc) and could use some copy editing.

(uwe)

Fix silly mistake in rev 1.12.

(thorpej)

It looks like Xen cpu_hatch() calls cpu_switchto() with prevlwp=NULL,
instead of calling idle_loop() directly.  I can't test a change to
cpu_hatch() right now so allow for prevlwp=NULL.

(ad)

- Fix an inverted test that could have prevented LWPs running on helper
  CPUs to teleport somewhere better during preempt().

- Fix a comment.

(ad)

Add missing El. Fix grammar.

(wiz)

Clean up dash use. Remove superfluous Pp.

(wiz)

It looks like Xen cpu_hatch() calls cpu_switchto() with prevlwp=NULL,
instead of calling idle_loop() directly.  I can't test a change to
cpu_hatch() right now so allow for prevlwp=NULL.

(ad)

Fix debug printf formats.

(martin)

Bump date in dlfcn(3)

(kamil)

Bump date in dlinfo(3)

(kamil)

Make the per-directory lock a rwlock.

(ad)

Cross-Link from dlfcn(3) to dlinfo(3)

(kamil)

Document dlinfo(3)

Man-page taken from FreeBSD.

(kamil)

Typo in comment.  s/LABELLOFFSET/LABELOFFSET/

(isaki)

more diag for cpu topology debug kernels:

- print the slow/fast status of each cpu
- if cpu topology is bogus, print info about the two CPUs that
  appear to have the same info.

(mrg)

Remove now unused mdlwp fields md_gc_pmap and md_gc_ptp.

(ad)

Fix a loltastic typo.

(ad)

apply FreeBSD revs r214848 and r214849:

    r214849 | jkim | 2010-11-05 13:24:26 -0700 (Fri, 05 Nov 2010) | 2 lines

    Add a forgotten change from the previous commit.

    r214848 | jkim | 2010-11-05 12:50:09 -0700 (Fri, 05 Nov 2010) | 13 lines

    Fix a use-after-free bug for extended IRQ resource[1].  When _PRS buffer is
    copied as a template for _SRS, a string pointer for descriptor name is also
    copied and it becomes stale as soon as it gets de-allocated[2].  Now _CRS is
    used as a template for _SRS as ACPI specification suggests if it is usable.
    The template from _PRS is still utilized but only when _CRS is not available
    or broken.  To avoid use-after-free the problem in this case, however, only
    mandatory fields are copied, optional data is removed, and structure length
    is adjusted accordingly.

    Reported by:    hps[1]
    Analyzed by:    avg[2]
    Tested by:      hps

This also fixes reading past the end of a structure as detected by KASAN.

(chs)

Improve diagnostic messages:
- Pass an additional argument to gttwsi_wait() to indicate what's
  going on, and report that, along with the error code from
  cv_timedwait(), if a timeout occurs.
- In gttwsi_send_stop(), if we don't get the expected NRS status,
  report which status we *did* get when a timeout occurs.

(thorpej)

Start to switch from OpenBSD timeout(9) API to callout(9) and workqueue(9)
Replace functions with NetBSD counterparts.

(sevan)

A final set of scheduler tweaks:

- Try hard to keep vfork() parent and child on the same CPU until execve(),
  failing that on the same core, but in all other cases scatter new LWPs
  among the different CPU packages, round robin, to try and get the best out
  of the available cache and bus bandwidth.

- Remove attempts at balancing.  Replace with a rate-limited skim of other
  CPU's run queues in sched_idle(), starting in the current package and
  moving outwards.  Add a sysctl tunable to change the interval.

- Make the cacheht_time tuneable take a milliseconds value.

- It's possible to configure things such that there's no CPU allowed to run
  an LWP.  Defeat this by always having a default:

Reported-by: syzbot+46968944dd9359ab93bc@syzkaller.appspotmail.com
Reported-by: syzbot+7f750a4cc230d1e831f9@syzkaller.appspotmail.com
Reported-by: syzbot+88d7675158f5cb4684db@syzkaller.appspotmail.com
Reported-by: syzbot+d409c2338150e9a8ae1e@syzkaller.appspotmail.com
Reported-by: syzbot+e152dc5bff188f67358a@syzkaller.appspotmail.com

(ad)

Enable sun8icrypto in GENERIC64.

But set its entropy estimate to zero until we have a better idea of
how it works.  Can't really hurt this way unless sun8icrypto is
maliciously dependent on all other inputs to the entropy pool, which
seems unlikely.

(riastradh)

add Makefile

(christos)

Make pri_t a short and get back some more space in struct lwp.

(ad)

NetBSD 9.99.37 - struct cpu_info & struct lwp changed.

(ad)

Handle HAVE_NBTOOL_CONFIG_H

(christos)

make this compile.

(christos)

Update callout_halt() to match src/sys/sys/callout.h

(sevan)

fix -Wsign-compare because module tests trigger it.

(christos)

initialize radix_tree_node_cache with PR_LARGECACHE

this increases the cache groups from 15 to 63 items in order
to reduce traffic between pool cache layers
this is the same as for other highly frequented pool caches as the pvpool and anonpool

(para)

Fix some more places in this file it's assumed cpu_info is smaller than 1kB.

(ad)

Add file completion.

(christos)

- Shuffle some items around in struct lwp to save space.  Remove an unused
  item or two.

- For lockstat, get a useful callsite for vnode locks (caller to vn_lock()).

(ad)

remove unused

(christos)

Tidy up the vnode locking around execve() on ELF images to acquire and
release the locks fewer times.  Proposed on tech-kern a very long time go.

(ad)

vput(): don't drop the vnode lock, carry the hold over into vrelel() which
might need it anyway.

(ad)

Clean up gttwsi's register access stuff:
- Garbage-collect the obsolete GTTWSI_ALLWINNER option; it hasn't been
  needed since FDT'ization of the Allwinner support code.
- Redefine thw "TWSI_*" register definitions to clearly call out:
  -> The Marvell flavor of the offsets
  -> The Allwinner flavor of the offsets
  ...and make the regular definitions indices into a register map.
- Pass the appropriate register map from the front-end to the core.
- Remove the customer register read/write callbacks -- they are no longer
  needed now that each front-end passes an appropriate register map to
  the core.

(thorpej)

- uvm_unmap_remove(): need to call pmap_update() with the object still
  locked, otherwise the page could gain a new identity and still be visible
  via a stale mapping.

- Adjust reference counts with atomics.

(ad)

re-enable BSDTAR now that the two issues (symlinks/atomicity) have been fixed.

(christos)

Leave pre-existing symlinks alone on extraction

When libarchive encounters an existing symbolic link during extraction
it removes that symbolic link first before overwriting it, unless
it is told that it can trust symlinks from the archive.

Placing symbolic links on known paths in the extracting subdirectory
is a simple way that a system administrator can place data at a
different location without having the overhead of a mountpoint.

Trusting symlinks from an archive is never safe because they can
maliciously overwrite files outside of the extraction directory.

This patch adds a linked-list to track of the symbolic links that
were created during extraction so that it does not trust them. This
way during extraction, libarchive can remove the symlinks it created,
but leave the pre-existing ones alone.

Unit-tests were adjusted for this new behavior.

(this is pull request 1300)

(christos)

Introduce ARCHIVE_EXTRACT_ATOMIC and set it by default on bsdtar.

This flag changes the way that regular files are extracted:

Instead of removing existing files first and re-creating them in
order to replace their contents, a temporary file is created and
when writing to the temporary file is completed, the file is
rename(2)d to the final destination name.

This has the effect of presenting a consistent view of the file to
the system (either the file with the new contents or the file with
the old contents). Removing and overwriting the file has the
undesired side effect that the the system can either not see the
file at all (from the time it is being removed till the time it is
being re-created), or worse it can see partial file contents. This
is problematic when extracting system files (for example shared
libraries).

Perhaps there should be a flag to disable it, when for example it
is not desirable because of space constraints, but then again
one can specify to unlink the file before.

(this is pull request 1289)

(christos)

Fix broken RCS NetBSD tag.

(isaki)

Redo previous using COHERENCY_UNIT.

(ad)

cpu topology:

- Fix a stupid bug where it complained about non-existent SMT on assymetric
  systems.

- Give each CPU a pointer back to the first CPU in same package.

(ad)

Bump MIN_LWP_ALIGNMENT to 64.

(ad)

KNF and misc whitespace cleanup.

(tsutsui)

Remove some unneeded kernel_lock handling.

(ad)

Nothing uses l->l_sleeperr any more.

(ad)

x86 pmap:

- It turns out that every page the pmap frees is necessarily zeroed.  Tell
  the VM system about this and use the pmap as a source of pre-zeroed pages.

- Redo deferred freeing of PTPs more elegantly, including the integration with
  pmap_remove_all().  This fixes problems with nvmm, and possibly also a crash
  discovered during fuzzing.

Reported-by: syzbot+a97186518c84f1d85c0c@syzkaller.appspotmail.com

(ad)

l->l_emap_gen isn't used any more.

(ad)

provide some semblance of valid cpu topology for big.little systems.

while attaching cpus, if the FDT provides "capacity-dmips-mhz" track
the fastest set, and call cpu_topology_set() with slow=true for any
cpus that are not the fastest.

bug fix for cpu_topology_set(): actually set ci_is_slow for slow cpus.

with this change, and -current's recent scheduler changes, this means
that long running processes run on the faster cores.  on RK3399 based
systems, i am seeing 20-50% speed ups for many tasks.

XXX: all this can be made common with armv7 big.little.

(mrg)

Revert "Compile with -O2 by default" change in previous rev 1.11.

COPTS is defined properly after sys/arch/arm/conf/Makefile.arm rev 1.53.
See discussion in source-changes-d@ for details:
https://mail-index.netbsd.org/source-changes-d/2020/01/thread1.html#011984

(tsutsui)

Fix "[: SMALL test, no fallback usage" error on miniroot installation.

Avoid and replace use of '-a', '(' and ')' operatos marked obsolescent
by modern POSIX.1-2017:
https://pubs.opengroup.org/onlinepubs/9699919799/utilities/test.html#tag_20_128_16
as suggested by kre@ in PR/54835.

Should be pulled up to netbsd-9.

(tsutsui)

PR/54853: Greg Oster: unable to 'unset filec' or 'unset edit' in csh
While here allow set edit=vi

(christos)

Fix miniroot installation failure on network configuration.  PR/54833

No particular comment in the PR.
Should be pulled up to netbsd-9.

(tsutsui)

Support imxuart_freq=0 (use existing settings)

(jmcneill)

In gttwsi_wait(), don't use cv_timedwait_sig() -- there's no reason to
catch signals here, and the code would need to handle them properly, which
this code does not.

(thorpej)

Note some more layouts that work with KB_SWAPCTRLCAPS.

(nia)

Support the combination of KB_SWAPCTRLCAPS and UK layout.

(nia)

Start porting to NetBSD. (WIP, incomplete)

(sevan)

Import rge(4) from OpenBSD for Realtek's RTL8125 based 2.5GbE network cards.

(sevan)

- uvm_pagezerocheck(): put a global lock around it to protect the single
  page mapping (DEBUG only).

- uvm_pagefree(): increment zeropages as needed.

(ad)

Define the min and max page size supported for the benefit of jemalloc

(christos)

Don't add the 2 reserved clusters before we determine if we using fat16/fat32.
From FreeBSD: https://reviews.freebsd.org/D23082:

Correct off-by-two issue when determining FAT type.

In the code we used NumClusters as the upper (non-inclusive) boundary
of valid cluster number, so the actual value was 2 (CLUST_FIRST) more
than the real number of clusters. This causes a FAT16 media with
65524 clusters be treated as FAT32 and might affect FAT12 media with
4084 clusters as well.

To fix this, we increment NumClusters by CLUST_FIRST after the type
determination.

(christos)

Simplify async_mixer handling.
- It makes FIOASYNC code in mixer_ioctl() symmetric.
- For readability, mixer_async_{add,remove}() should take pid argument
  though pid is always curproc.

(isaki)

Remove old debug #ifdefs.

(isaki)

UVM_PAGE_TREE_PENALTY isn't used any more.

(ad)

use getline(3)

(christos)

Add WP

From Rocky Hotas, thanks!

(leot)

Fix markup typo.

(maya)

Add some more licenses appearing in external.
Remove mentions of src/gnu that no longer exists.
Mention /usr/pkg/etc.
Bump date

(maya)

When asked to create a whole-disk partition for unknown usage, do not only
assert(false), but actually return an error (in case "assert" is a nop).

(martin)

Use first-level heading for "NetBSD" title in README.md

(nia)