Pass B_PHYS when reading from device.  Xbd(4) at least checks
this flag and may trigger an assertion.

(hannken)

Tickets #123 and #124

(martin)

Pull up following revision(s) (requested by hannken in ticket #124):

tests/lib/libc/sys/t_ptrace_syscall_wait.h: revision 1.3
sys/arch/sparc/sparc/syscall.c: revision 1.32
sys/arch/sparc/sparc/vm_machdep.c: revision 1.108

Adjust pc/npc before syscall allowing EJUSTRETURN to return
to the next instruction.  Only ERESTART should return to
the same instruction.  Differences to sparc64 reduced.

Test t_ptrace_wait:syscallemu1 now passes on sparc.

Fixes PR kern/52166 "syscallemu does not work on sparc (32-bit)"

Ok: Martin Husemann

(martin)

Remove directory, requested by mrg in ticket #123:

external/gpl3/gcc.old (remove)

This directory is not used on this branch.

(martin)

Ignore non-recoverable and critical limits smaller than the warning limits.
These are usually invalid.

Name the limit flags to make code more readable.

(mlelstv)

in6: make sure a user-specified checksum field is within a packet

From OpenBSD

(ozaki-r)

octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

  The entropy is provided by the jitter of 125 of 128 free-running
  oscillators XORed into a 128-bit LFSR.  The LFSR accumulates entropy
  over 81 cycles, after which it is fed into a SHA-1 engine.
  [...]
  The SHA-1 engine runs once every 81 cycles.
  [...]
  The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

(riastradh)

Merge between
        src-external-gpl3-gdb-20200913-1743: 1.1.1.1
and
        src-external-gpl3-gdb-20230321-1235: 1.1.1.1

(christos)

PR/57279: Izumi Tsutsui: Fix some {int,long} -> time_t. Still things will
break eventually because parts of the nfs protocol assume time_t will fit
in 32 bits.

(christos)

Use "sigjmp_buf loc" after switch to sigsetjmp()/siglongjmp().

Fixes errors and aborts on sparc at least.

(hannken)

Ticket #122

(martin)

Pull up following revision(s) (requested by mrg in ticket #122):

sys/external/bsd/drm2/dist/drm/amd/amdgpu/amdgpu_gart.c: revision 1.11
sys/external/bsd/drm2/nouveau/nouveau_pci.c: revision 1.37
sys/external/bsd/drm2/radeon/radeon_pci.c: revision 1.22
sys/dev/wscons/wsdisplay.c: revision 1.166
sys/dev/wscons/wsdisplayvar.h: revision 1.57

amdgpu: Fix bogus loop invariant assertions in amdgpu_gart_map.
nouveau: Kick out genfb on firmware framebuffer before initializing.

PR kern/53126

radeon: Kick out genfb on firmware framebuffer before initializing.
this is the same change as nouveau_pci.c:1.37, and should fix at
least PR#56714 and i thought at least another PR i can't find right
now.  it fixes at least 2 different radeon cards for me on UEFI
booted system.

(martin)

Adjust pc/npc before syscall allowing EJUSTRETURN to return
to the next instruction.  Only ERESTART should return to
the same instruction.  Differences to sparc64 reduced.

Test t_ptrace_wait:syscallemu1 now passes on sparc.

Fixes PR kern/52166 "syscallemu does not work on sparc (32-bit)"

Ok: Martin Husemann

(hannken)

__HAVE_HASHLOCKED_ATOMICS needs to be visible to userland

(martin)

in6: reject setting negative values but -1 via setsockopt(IPV6_CHECKSUM)

Same as OpenBSD.

(ozaki-r)

merge xbitmaps 1.1.3 and xorg-cf-files 1.0.8.

also, fix riscv64 on netbsd to define _XSERVER64.

(mrg)

merge new font packages.

(mrg)

Do a better job handling EACCES errors from exec() calls.  If the
EACCES is from the namei(), treat it just like ENOENT or ENOTDIR
(and if that is the final error, the exit status from a failed exec
will be 127).  If the EACCES is from the exec() itself, that indicates
the file to be run exists, but has no 'x' permission.  That's a
meaningful error (as distinct from just "yet another PATH element
search failure").

While here, return the first meaingful error we encountered while
searching PATH, rather than the last (and ENOENT if there are none
of those).

This change results in some failed command executions returning status
127 now, where they returned 126 before - which better reflects the
intent of those values (127 is simply "not found" whereas 126 is "found
but couldn't be executed").

We still do nothing to distinguish errors encountered looking up the
command name give, with errors encountered (by the kernel) attempting to
run an interpreter needed for the exec to succeed (#! line path, or
/libexec/ld.elf_so and similar - or anything else of a similar nature).

(kre)

Switch from using _setjmp()/_longjmp() (on BSD systems which aren't SVR4)
(and setjmp()/longjmp() elsewhere) to using sigsetjmp()/siglongjmp()
everywhere.

NFCI.

(kre)

Change a few #defines from octal to hex (pdp11 days are long gone).
Improve the layout of those definitions at the same time.

NFC.

(kre)

PR misc/57276 from Nan Xiao

Remove redundant word ("to").

(kre)

make: handle .PHONY consitently on interrupt

JobDeleteTarget skips .PHONY targets
CompatDeleteTarget should do the same

This addresses https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=269663

(sjg)

Can't use disklabel to find the real size of a disk since a fake disklabel
read from disk is trusted. Revert to use drvctl again.

(mlelstv)

s/Brigthness/Brightness/ in comment.

(andvar)

s/enougth/enough/

(andvar)

Avoid overflow of nfs_commitsize on machines with > 32GB RAM.

(mlelstv)

openssh 9.3 is out

(wiz)

Pull up following revision(s) (requested by hannken in ticket #121):

sys/dev/vnd.c: revision 1.288

Do not limit the number of pending requests for the worker thread.

With wedge on vnd it prevents a deadlock when requests get queued with
biodone() -> dkstart() -> vndstrategy().

Fixes PR kern/57263 "vnd locks up when using vn_rdwr"

(martin)

mountd(8): comma before "and"

Not a serial comma, technically, as there are only two items, but it
makes reading easier, IMHO.

(uwe)

mountd(8): fix minor markup nits

While here, use $() instead of `` in the example.

(uwe)

Ticket #1614

(martin)

Pull up following revision(s) (requested by hannken in ticket #1614):

sys/dev/vnd.c: revision 1.288

Do not limit the number of pending requests for the worker thread.

With wedge on vnd it prevents a deadlock when requests get queued with
biodone() -> dkstart() -> vndstrategy().

Fixes PR kern/57263 "vnd locks up when using vn_rdwr"

(martin)

Ticket #121

(martin)

+BQP    bounded-error quantum polynomial time
+KEM    key encapsulation mechanism
+PQC    post-quantum cryptography

(jschauma)

service(8): fix markup

(uwe)

Add SPIR (standard portable intermediate representation).

(fcambus)

Tickets #118 - #120, fix pasto in #117

(martin)

Pull up following revision(s) (requested by nros in ticket #120):

external/mit/xorg/lib/xkeyboard-config/xkeyboard-config.pc.in: revision 1.2
external/mit/xorg/lib/xkeyboard-config/xkeyboard-config.man: revision 1.12

Fix xkb_base in the pkg-config file and mapage for xkeyboard-config

The xkbdata files are installed in /usr/X11R7/lib/X11/xkb
not /usr/X11R7/share/X11/xkb. The wrong directory causes
build failures in pkgsrc packages that use the xkb_base
variable.

Also base the datadir on the vars from bsd.x11.mk.

Also use @xkb_base@ instead of /usr/X11R7/share/lib/X11/xkb
in the manpage to show the right xkbdata directory.

(martin)

Tickets #1612 and #1613

(martin)

Pull up following revision(s) (requested by kardel in ticket #119):

sys/net/npf/npf_mbuf.c: revision 1.25
sys/net/npf/npf.h: revision 1.64
sys/net/npf/npf_sendpkt.c: revision 1.23

PR kern/56052:
allow block-return packets passed through without rule matching.
Included up-stream ashttps://github.com/rmind/npf/pull/115

(martin)

Pull up following revision(s) (requested by kardel in ticket #119):

sys/net/npf/npf_mbuf.c: revision 1.25
sys/net/npf/npf.h: revision 1.64
sys/net/npf/npf_sendpkt.c: revision 1.23

PR kern/56052:
allow block-return packets passed through without rule matching.
Included up-stream ashttps://github.com/rmind/npf/pull/115

(martin)

Pull up following revision(s) (requested by kardel in ticket #1612):

usr.sbin/npf/npf-params.7: revision 1.9
sys/net/npf/npf_params.c: revision 1.6

PR kern/55654:

Switch default for parameter npf ip4.reassembly to 1.

This makes the NPF default configuration comply with host
requirements for IPv4.

(martin)

Pull up following revision(s) (requested by kardel in ticket #118):

usr.sbin/npf/npf-params.7: revision 1.9
sys/net/npf/npf_params.c: revision 1.6

PR kern/55654:

Switch default for parameter npf ip4.reassembly to 1.

This makes the NPF default configuration comply with host
requirements for IPv4.

(martin)

Do not limit the number of pending requests for the worker thread.

With wedge on vnd it prevents a deadlock when requests get queued with
biodone() -> dkstart() -> vndstrategy().

Fixes PR kern/57263 "vnd locks up when using vn_rdwr"

(hannken)

In bwrite() if the write() returned -1, the associated errno was
being used after a (possible) call to printf (only happens when
debugging) and a certain call to free() (via brelse()).
Make a copy, and use that instead.

(kre)

Correct a minor word usage error in a comment.  NFC.

(kre)

Use %zd to print ssize_t (%z[uxo] for size_t), rather than %jd
(just like the other adjacent printfs do)

Should fix i386 build.

(kre)

Don't inappropriately use xargs to run rcorder -- xargs is designed
to run the given command as many times as necessary, but rcorder must
be run just once, with all the rc.d scripts as args.  If that turns out
to be too many (E2BIG on exec) then we have more serious problems, working
around it by use of xargs just makes a mess (for rcorder).

Make the -e option cause the script to exit 0 if some enabled script was
found, and 1 if not, so one can do:
if service -e foobar >/dev/null
then
whatever we want to do when foobar is enabled
else
anything we want instead if it is not
fi
Someday perhaps add a -q to suppress the output instead of requiring
the redirect, but that day is not this one.  Make all error exit status
codes be >1 (and use the standard 126 instead of 255, for "exec failed").

Note that with more than one service given with -e, or when none are (so
all enabled scripts are listed) the exit status is 0 if any enabled script
was found (ie: anything was printed to stdout), not if all named scripts
exist and are enabled.  Testing the status works best with just one
script name given as an arg.

Correctly quote script names (and use eval where required) so scripts
(or paths to scripts) containing spaces, tabs, or shell operator chars
might be handled correctly (newlines in names will still cause failures.)

Treat rc.d scripts that do not have an rcvar (hence are always enabled)
as enabled scripts, rather than disabled - but ignore the LOGIN DAEMON...
pseudo-scripts.

While here, replace archaic uses of "test" ('[') (that is, -a etc) with
standard defined usages only, replace a baroque use of sed with sh code
(which is simpler and much faster, not that speed matters) - but do use sed
instead of two grep invocations.  Replace all uses of "echo" with "printf"
(just "because"!)  Make the usage more useful (explain what the various
option combinations achieve, explicitly), and also while here, make the
formatting look closer to something I can deal with (personally I prefer
tab indents almost everywhere, but 4 space is OK...)  Aligned runs of 8
spaces were all replaced by a tab.  For the options, use sh boolean cmds
(true|false), and simply run them, rather than making them be empty or set
and using test -n, it is easier to follow (and a tiny fraction of a ns faster).
Change a comment so what it says is relevant to the code that is present,
rather than to the change (referring to code that used to be present) with
which it was added.

Catch the manual page up with the minor parts of this intended to be
visible to users (like the exit status change).

(kre)

Add (newly added): expm1l log1pl log2l lgammal tgammal
so that the library gets the correct symbols defined in it.

(kre)

atf-c-api.3: correct ATF_TP_ADD_TC() examples to use "tp"

(gutteridge)

Don't brelse() if bread() fails. The kernel does this for us. Our bread()
implementation just exits on failure, but if it didn't we would double-free.
From Ed Maste (https://reviews.freebsd.org/D39069)

(christos)

modernize; release bpp and set it to NULL as the kernel does even if we
never return here.

(christos)

Ticket #117

(jdc)

Pull up the following revision (requested by martin in ticket #117):

build.sh: revision 1.366 via patch

Avoid the dependency on a populated tooldir (or building the tools)
when simply doing mkrepro-timestamp and the current repository setups
does not actually require it.

(jdc)

Ticket #1611

(jdc)

Pull up the following revision (requested by martin in ticket #1611):

build.sh: revision 1.366 via patch

Avoid the dependency on a populated tooldir (or building the tools)
when simply doing mkrepro-timestamp and the current repository setups
does not actually require it.

(jdc)

Ticket #1807

(jdc)

Pull up the following revision (requested by martin in ticket #1807):

build.sh: revision 1.366 via patch

Avoid the dependency on a populated tooldir (or building the tools)
when simply doing mkrepro-timestamp and the current repository setups
does not actually require it.

(jdc)

Ticket #1806

(jdc)

libm: Need <math.h> for __HAVE_LONG_DOUBLE.

Fixes missing definitions of lrintl, llrintl on various platforms.

(riastradh)

libm: Fill in more dummy long double transcendental functions.

This should cover everything from C99.

(riastradh)

vn_open(9): Add assertion that vp is locked on return.

Null out vp internally out of paranoia so we'll crash in evaluating
the assertion if we ever reach it via one of the vput paths.

(riastradh)

vn_open(9): Clarify that this returns a locked vnode.

Comment only, no functional change intended.

(riastradh)

Avoid the dependency on a populated tooldir (or building the tools)
when simply doing mkrepro-timestamp and the current repository setups
does not actually require it.

(martin)

Ticket #116

(martin)

Pull up following revision(s) (requested by msaitoh in ticket #116):

sys/dev/acpi/igpio_acpi.c: revision 1.2
share/man/man4/igpio.4: revision 1.4

igpio(4): Disable Baytrail and Lynxpoint support. Fixes PR kern/57225.
It seems additional code is required to support Baytrail and Lynxpoint.

Disable those support.

(martin)

PR 57241: switch mips64el back to old binutils for now

(martin)

Note more user visible changes by me that will appear in NetBSD 10.0.

(tsutsui)

Move changes pulled up to netbsd-10 from CHANGES to CHANGES.prev.

(tsutsui)

make: document how read-only variables can be made read-write

(rillig)

make: fix memory leak when parsing ':from=to' modifiers

Since var.c 1.925 from 2021-04-12.

(rillig)

simplify previous

(christos)

remove "nouveau" from a comment.  noted by jmcneill.

(mrg)

Pull up following revision(s) (requested by riastradh in ticket #1806):

sys/kern/vfs_syscalls.c: revision 1.557

open(2): Don't map ERESTART to EINTR.

If a file or device's open function returns ERESTART, respect that --
restart the syscall; don't pretend a signal has been delivered when
it was not.  If an SA_RESTART signal was delivered, POSIX does not
allow it to fail with EINTR:

    SA_RESTART
        This flag affects the behavior of interruptible functions;
        that is, those specified to fail with errno set to [EINTR].
        If set, and a function specified as interruptible is
        interrupted by this signal, the function shall restart and
        shall not fail with [EINTR] unless otherwise specified.  If
        an interruptible function which uses a timeout is restarted,
        the duration of the timeout following the restart is set to
        an unspecified value that does not exceed the original
        timeout value.  If the flag is not set, interruptible
        functions interrupted by this signal shall fail with errno
        set to [EINTR].

https://pubs.opengroup.org/onlinepubs/9699919799/functions/sigaction.html

Nothing in the POSIX definition of open specifies otherwise.

In 1990, Kirk McKusick added these lines with a mysterious commit
message:
Author: Kirk McKusick <mckusick>
Date:  Tue Apr 10 19:36:33 1990 -0800
    eliminate longjmp from the kernel (for karels)
diff --git a/sys/kern/vfs_syscalls.c b/sys/kern/vfs_syscalls.c
index 7bc7b39bbf..d572d3a32d 100644
--- a/sys/kern/vfs_syscalls.c
+++ b/sys/kern/vfs_syscalls.c
@@ -14,7 +14,7 @@
  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  *
- *    @(#)vfs_syscalls.c      7.42 (Berkeley) 3/26/90
+ *    @(#)vfs_syscalls.c      7.43 (Berkeley) 4/10/90
  */
#include "param.h"
@@ -530,8 +530,10 @@ copen(scp, fmode, cmode, ndp, resultfd)
        if (error = vn_open(ndp, fmode, (cmode & 07777) &~ S_ISVTX)) {
                crfree(fp->f_cred);
                fp->f_count--;
-              if (error == -1)        /* XXX from fdopen */
-                      return (0);    /* XXX from fdopen */
+              if (error == EJUSTRETURN)      /* XXX from fdopen */
+                      return (0);            /* XXX from fdopen */
+              if (error == ERESTART)
+                      error = EINTR;
                scp->sc_ofile[indx] = NULL;
                return (error);
        }

(found via this git import of the CSRG history:
https://github.com/robohack/ucb-csrg-bsd/commit/cce2869b7ae5d360921eb411005b328a29c4a3fe

This change appears to have served two related purposes:
1. The fdopen function (the erstwhile open routine for /dev/fd/N)
  used to return -1 as a hack to mean it had just duplicated the fd;
  it was recently changed by Mike Karels, in kern_descrip.c 7.9, to
  return EJUSTRETURN, now defined to be -2, presumably to avoid a
  conflict with ERESTART, defined to be -1.  So this change finished
  part of the change by Mike Karels to use a different magic return
  code from fdopen.
  Of course, today we use still another disgusting hack, EDUPFD, for
  the same purpose, so none of this is relevant any more.
2. Prior to April 1990, the kernel handled signals during tsleep(9)
  by longjmping out to the system call entry point or similar.  In
  April 1990, Mike Karels worked to convert all of that into
  explicit unwind logic by passing through EINTR or ERESTART as
  appropriate, instead of setjmp at each entry point.

However, it's not clear to me why this setjmp/longjmp and
fdopen/-1/EJUSTRETURN renovation justifies unconditional logic to map
ERESTART to EINTR in open(2).  I suspect it was a mistake.

In 2013, the corresponding logic to map ERESTART to EINTR in open(2)
was removed from FreeBSD:

  r246472 | kib | 2013-02-07 14:53:33 +0000 (Thu, 07 Feb 2013) | 11 lines
  Stop translating the ERESTART error from the open(2) into EINTR.
  Posix requires that open(2) is restartable for SA_RESTART.
  For non-posix objects, in particular, devfs nodes, still disable
  automatic restart of the opens. The open call to a driver could have
  significant side effects for the hardware.
  Noted and reviewed by:  jilles
  Discussed with: bde
  MFC after:      2 weeks

Index: vfs_syscalls.c
===================================================================
--- vfs_syscalls.c      (revision 246471)
+++ vfs_syscalls.c      (revision 246472)
@@ -1106,8 +1106,6 @@
                                goto success;
                }
-              if (error == ERESTART)
-                      error = EINTR;
                goto bad;
        }
        td->td_dupfd = 0;

https://cgit.freebsd.org/src/commit/sys/kern/vfs_syscalls.c?id=2ca49983425886121b506cb5126b60a705afc38c">https://cgit.freebsd.org/src/commit/sys/kern/vfs_syscalls.c?id=2ca49983425886121b506cb5126b60a705afc38c

It's not clear to me that there's any reason to treat device nodes
specially here; in fact, if a driver's .d_open routine sleeps and is
woken by a concurrent revoke without a signal pending or with an
SA_RESTART signal pending, it is wrong for it to fail with EINTR.

But it MUST restart the whole system call rather than continue
sleeping in a loop or just exit the loop and continue to open,
because it is mandatory in the security model of revoke for open(2)
to retry the permissions check at that point.

PR kern/57260

(martin)

Ticket #161

(martin)

Pull up following revision(s) (requested by riastradh in ticket #1610):

sys/kern/vfs_syscalls.c: revision 1.557

open(2): Don't map ERESTART to EINTR.

If a file or device's open function returns ERESTART, respect that --
restart the syscall; don't pretend a signal has been delivered when
it was not.  If an SA_RESTART signal was delivered, POSIX does not
allow it to fail with EINTR:

    SA_RESTART
        This flag affects the behavior of interruptible functions;
        that is, those specified to fail with errno set to [EINTR].
        If set, and a function specified as interruptible is
        interrupted by this signal, the function shall restart and
        shall not fail with [EINTR] unless otherwise specified.  If
        an interruptible function which uses a timeout is restarted,
        the duration of the timeout following the restart is set to
        an unspecified value that does not exceed the original
        timeout value.  If the flag is not set, interruptible
        functions interrupted by this signal shall fail with errno
        set to [EINTR].

https://pubs.opengroup.org/onlinepubs/9699919799/functions/sigaction.html

Nothing in the POSIX definition of open specifies otherwise.

In 1990, Kirk McKusick added these lines with a mysterious commit
message:
Author: Kirk McKusick <mckusick>
Date:  Tue Apr 10 19:36:33 1990 -0800
    eliminate longjmp from the kernel (for karels)
diff --git a/sys/kern/vfs_syscalls.c b/sys/kern/vfs_syscalls.c
index 7bc7b39bbf..d572d3a32d 100644
--- a/sys/kern/vfs_syscalls.c
+++ b/sys/kern/vfs_syscalls.c
@@ -14,7 +14,7 @@
  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  *
- *    @(#)vfs_syscalls.c      7.42 (Berkeley) 3/26/90
+ *    @(#)vfs_syscalls.c      7.43 (Berkeley) 4/10/90
  */
#include "param.h"
@@ -530,8 +530,10 @@ copen(scp, fmode, cmode, ndp, resultfd)
        if (error = vn_open(ndp, fmode, (cmode & 07777) &~ S_ISVTX)) {
                crfree(fp->f_cred);
                fp->f_count--;
-              if (error == -1)        /* XXX from fdopen */
-                      return (0);    /* XXX from fdopen */
+              if (error == EJUSTRETURN)      /* XXX from fdopen */
+                      return (0);            /* XXX from fdopen */
+              if (error == ERESTART)
+                      error = EINTR;
                scp->sc_ofile[indx] = NULL;
                return (error);
        }

(found via this git import of the CSRG history:
https://github.com/robohack/ucb-csrg-bsd/commit/cce2869b7ae5d360921eb411005b328a29c4a3fe

This change appears to have served two related purposes:
1. The fdopen function (the erstwhile open routine for /dev/fd/N)
  used to return -1 as a hack to mean it had just duplicated the fd;
  it was recently changed by Mike Karels, in kern_descrip.c 7.9, to
  return EJUSTRETURN, now defined to be -2, presumably to avoid a
  conflict with ERESTART, defined to be -1.  So this change finished
  part of the change by Mike Karels to use a different magic return
  code from fdopen.
  Of course, today we use still another disgusting hack, EDUPFD, for
  the same purpose, so none of this is relevant any more.
2. Prior to April 1990, the kernel handled signals during tsleep(9)
  by longjmping out to the system call entry point or similar.  In
  April 1990, Mike Karels worked to convert all of that into
  explicit unwind logic by passing through EINTR or ERESTART as
  appropriate, instead of setjmp at each entry point.

However, it's not clear to me why this setjmp/longjmp and
fdopen/-1/EJUSTRETURN renovation justifies unconditional logic to map
ERESTART to EINTR in open(2).  I suspect it was a mistake.

In 2013, the corresponding logic to map ERESTART to EINTR in open(2)
was removed from FreeBSD:

  r246472 | kib | 2013-02-07 14:53:33 +0000 (Thu, 07 Feb 2013) | 11 lines
  Stop translating the ERESTART error from the open(2) into EINTR.
  Posix requires that open(2) is restartable for SA_RESTART.
  For non-posix objects, in particular, devfs nodes, still disable
  automatic restart of the opens. The open call to a driver could have
  significant side effects for the hardware.
  Noted and reviewed by:  jilles
  Discussed with: bde
  MFC after:      2 weeks

Index: vfs_syscalls.c
===================================================================
--- vfs_syscalls.c      (revision 246471)
+++ vfs_syscalls.c      (revision 246472)
@@ -1106,8 +1106,6 @@
                                goto success;
                }
-              if (error == ERESTART)
-                      error = EINTR;
                goto bad;
        }
        td->td_dupfd = 0;

https://cgit.freebsd.org/src/commit/sys/kern/vfs_syscalls.c?id=2ca49983425886121b506cb5126b60a705afc38c">https://cgit.freebsd.org/src/commit/sys/kern/vfs_syscalls.c?id=2ca49983425886121b506cb5126b60a705afc38c

It's not clear to me that there's any reason to treat device nodes
specially here; in fact, if a driver's .d_open routine sleeps and is
woken by a concurrent revoke without a signal pending or with an
SA_RESTART signal pending, it is wrong for it to fail with EINTR.

But it MUST restart the whole system call rather than continue
sleeping in a loop or just exit the loop and continue to open,
because it is mandatory in the security model of revoke for open(2)
to retry the permissions check at that point.

PR kern/57260

(martin)

Tickets #113 - #115

(martin)

Pull up following revision(s) (requested by riastradh in ticket #115):

sys/kern/vfs_syscalls.c: revision 1.557

open(2): Don't map ERESTART to EINTR.

If a file or device's open function returns ERESTART, respect that --
restart the syscall; don't pretend a signal has been delivered when
it was not.  If an SA_RESTART signal was delivered, POSIX does not
allow it to fail with EINTR:

    SA_RESTART
        This flag affects the behavior of interruptible functions;
        that is, those specified to fail with errno set to [EINTR].
        If set, and a function specified as interruptible is
        interrupted by this signal, the function shall restart and
        shall not fail with [EINTR] unless otherwise specified.  If
        an interruptible function which uses a timeout is restarted,
        the duration of the timeout following the restart is set to
        an unspecified value that does not exceed the original
        timeout value.  If the flag is not set, interruptible
        functions interrupted by this signal shall fail with errno
        set to [EINTR].

https://pubs.opengroup.org/onlinepubs/9699919799/functions/sigaction.html

Nothing in the POSIX definition of open specifies otherwise.

In 1990, Kirk McKusick added these lines with a mysterious commit
message:
Author: Kirk McKusick <mckusick>
Date:  Tue Apr 10 19:36:33 1990 -0800
    eliminate longjmp from the kernel (for karels)
diff --git a/sys/kern/vfs_syscalls.c b/sys/kern/vfs_syscalls.c
index 7bc7b39bbf..d572d3a32d 100644
--- a/sys/kern/vfs_syscalls.c
+++ b/sys/kern/vfs_syscalls.c
@@ -14,7 +14,7 @@
  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  *
- *    @(#)vfs_syscalls.c      7.42 (Berkeley) 3/26/90
+ *    @(#)vfs_syscalls.c      7.43 (Berkeley) 4/10/90
  */
#include "param.h"
@@ -530,8 +530,10 @@ copen(scp, fmode, cmode, ndp, resultfd)
        if (error = vn_open(ndp, fmode, (cmode & 07777) &~ S_ISVTX)) {
                crfree(fp->f_cred);
                fp->f_count--;
-              if (error == -1)        /* XXX from fdopen */
-                      return (0);    /* XXX from fdopen */
+              if (error == EJUSTRETURN)      /* XXX from fdopen */
+                      return (0);            /* XXX from fdopen */
+              if (error == ERESTART)
+                      error = EINTR;
                scp->sc_ofile[indx] = NULL;
                return (error);
        }

(found via this git import of the CSRG history:
https://github.com/robohack/ucb-csrg-bsd/commit/cce2869b7ae5d360921eb411005b328a29c4a3fe

This change appears to have served two related purposes:
1. The fdopen function (the erstwhile open routine for /dev/fd/N)
  used to return -1 as a hack to mean it had just duplicated the fd;
  it was recently changed by Mike Karels, in kern_descrip.c 7.9, to
  return EJUSTRETURN, now defined to be -2, presumably to avoid a
  conflict with ERESTART, defined to be -1.  So this change finished
  part of the change by Mike Karels to use a different magic return
  code from fdopen.
  Of course, today we use still another disgusting hack, EDUPFD, for
  the same purpose, so none of this is relevant any more.
2. Prior to April 1990, the kernel handled signals during tsleep(9)
  by longjmping out to the system call entry point or similar.  In
  April 1990, Mike Karels worked to convert all of that into
  explicit unwind logic by passing through EINTR or ERESTART as
  appropriate, instead of setjmp at each entry point.

However, it's not clear to me why this setjmp/longjmp and
fdopen/-1/EJUSTRETURN renovation justifies unconditional logic to map
ERESTART to EINTR in open(2).  I suspect it was a mistake.

In 2013, the corresponding logic to map ERESTART to EINTR in open(2)
was removed from FreeBSD:

  r246472 | kib | 2013-02-07 14:53:33 +0000 (Thu, 07 Feb 2013) | 11 lines
  Stop translating the ERESTART error from the open(2) into EINTR.
  Posix requires that open(2) is restartable for SA_RESTART.
  For non-posix objects, in particular, devfs nodes, still disable
  automatic restart of the opens. The open call to a driver could have
  significant side effects for the hardware.
  Noted and reviewed by:  jilles
  Discussed with: bde
  MFC after:      2 weeks

Index: vfs_syscalls.c
===================================================================
--- vfs_syscalls.c      (revision 246471)
+++ vfs_syscalls.c      (revision 246472)
@@ -1106,8 +1106,6 @@
                                goto success;
                }
-              if (error == ERESTART)
-                      error = EINTR;
                goto bad;
        }
        td->td_dupfd = 0;

https://cgit.freebsd.org/src/commit/sys/kern/vfs_syscalls.c?id=2ca49983425886121b506cb5126b60a705afc38c">https://cgit.freebsd.org/src/commit/sys/kern/vfs_syscalls.c?id=2ca49983425886121b506cb5126b60a705afc38c

It's not clear to me that there's any reason to treat device nodes
specially here; in fact, if a driver's .d_open routine sleeps and is
woken by a concurrent revoke without a signal pending or with an
SA_RESTART signal pending, it is wrong for it to fail with EINTR.

But it MUST restart the whole system call rather than continue
sleeping in a loop or just exit the loop and continue to open,
because it is mandatory in the security model of revoke for open(2)
to retry the permissions check at that point.

PR kern/57260

(martin)

Pull up following revision(s) (requested by thorpej in ticket #114):

lib/libc/sys/open.2: revision 1.66

Document the error code when O_REGULAR is specified and the last
path component is not a regular file.

(martin)

Pull up following revision(s) (requested by thorpej in ticket #113):

sys/dev/usb/ucom.c: revision 1.137

In the HUP-wait path in ucomopen():
- Use cv_timedwait() rather than cv_timedwait_sig(); the wait here is
bounded (and fairly short besides) and seems appropriate to treat like
other uninterruptible waits.  The behavior is now consistent with com(4)
in this regard.
- Map EWOULDBLOCK return from cv_timedwait() to 0, as the successful passage
of time is not an error in this case.
- If the HUP-wait time has passed, clear the HUP-wait timestamp.

PR kern/57259 (although insufficient -- another change to vfs_syscalls.c
is required)

(martin)

This example is plain asm code, running ctf tools on it is not usefull

(martin)

radeon: Kick out genfb on firmware framebuffer before initializing.

this is the same change as nouveau_pci.c:1.37, and should fix at
least PR#56714 and i thought at least another PR i can't find right
now.  it fixes at least 2 different radeon cards for me on UEFI
booted system.

XXX: pullup-10 also include the original change:
  http://mail-index.netbsd.org/source-changes/2023/03/01/msg143606.html

(mrg)

Add:
HCC - Home Cable Computer
HCCA - Home Cable Computer Adaptor (yes, that is the original spelling)

(thorpej)

fix few typos in comments and log messages.

(andvar)

doc: ch-ch-changes

(nia)

Document __HAVE_UCAS_FULL and __HAVE_UCAS_MP.

(thorpej)

Adjust the tilde expansion test to deal with (very) recent changes
to what the shell produces for tilde expansions, when the expansion
of the tilde prefix ends with '/' and the character immediately following
is another '/' - previously /bin/sh (and most other shells, but not all)
retained both slashes, and this test expected that behaviour.

No longer, now only one of the two will appear.  Adjust the expected
test results accordingly, and add an extra loop iteration to make sure
this is thoroughly tested (one more tilde expansion value).

Also, add two new test cases that test for the new (explicit - though
it was always stated this way, but not as explicitly) that if HOME is
an empty string (not unset - that remains an unspecified case, as it
was) then the expansion of ~ must generate "", and not nothing.
The current test was unable to distinguish those two, since it
simply looked for characters in the output, so add a new test cases
explicitly to test for this particular case.  This also means (also
previously in the standard, but not as explicitly) that shells are
not permitted to decide "The ~ expansion produces an empty string, that
is weird, let's generate something else instead" which some shells were
doing.

(kre)

Adjust tilde expansion as will be documented in the forthcoming
version of the POSIX standard (Issue 8).  I believe we were already
compliant with what is to be required, but POSIX is now encouraging
(and will likely require in a later version) that if a tilde expansion
produces a string which ends in a '/' and the '~' that was expanded
is immediately followed by a '/' in the input word, that one of those
two slashes be omitted.  The worst (current) example of this is
when HOME=/ and we expand ~/foo - previously producing //foo which is
(in POSIX) a path with implementation defined semantics, and so not
what we should be generating by accident.  Change that, so now if
the ~ prefix expansion ends in a '/' and there is a '/' following
immediately after, the resulting word contains only one of those
chars (in the example just given, we will now produce /foo instead).

POSIX is also making it clear that the expansion that results from
the tilde expansion is treated as quoted (not subject to pathname
expansion, or field splitting, or any var/arith/command substitutions)
and that if HOME="" the expansion of ~ must generate "" (not nothing).
Our implementation did all of that already (though older versions
used to treat an empty expansion of HOME the same as if HOME was
unset - that was fixed some time ago).

The actual modification made here is probably smaller than this log entry,
and without added comments, certainly is!

(kre)

Unless -v is given, ignore EEXIST errors from the IOC_CPU_UCODE_APPLY ioctl()
used to implement "cpuctl ucode N",  which indicates that the microcode
to be loaded already exists in the CPU, and as such, isn't really a
very interesting "error".

(kre)

fix typos

(wiz)

portfeatures(9): the man page for the __HAVE macros

The man page itself is a skeleton/template for now.  Please add short
descriptions and xrefs.  If the target man page doesn't yet document
relevant __HAVE macros (hi, mutex(9)), please add something there,
possibly in a .Sh IMPLEMENTATION NOTES section.

(uwe)

ucom(4): Simplify logic fixing PR kern/57259.

cv_timedwait only ever returns 0 or EWOULDBLOCK, so this would always
return ERESTART anyway.

No functional change intended.

(riastradh)

In "devlist" mode, exit with a 0 return code if any nvme devices are
found, rather than exiting with 1 return code always.

(simonb)

mpc is updated to 1.3.1.

(mrg)

updates for mpc 1.3.1

(mrg)

note mpfr has a new version (4.2.0) and is updated, and
that mpc has a new version (1.3.1).

(mrg)

updates for mpfr 4.2.0.

(mrg)

merge mpfr 4.2.0.

(mrg)

Locate wedges as boot device and also match a partition GUID.

(mlelstv)

s/degub/debug/ in syspkgs definitions.

(andvar)

Document the error code when O_REGULAR is specified and the last
path component is not a regular file.

(thorpej)

open(2): Don't map ERESTART to EINTR.

If a file or device's open function returns ERESTART, respect that --
restart the syscall; don't pretend a signal has been delivered when
it was not.  If an SA_RESTART signal was delivered, POSIX does not
allow it to fail with EINTR:

    SA_RESTART
        This flag affects the behavior of interruptible functions;
        that is, those specified to fail with errno set to [EINTR].
        If set, and a function specified as interruptible is
        interrupted by this signal, the function shall restart and
        shall not fail with [EINTR] unless otherwise specified.  If
        an interruptible function which uses a timeout is restarted,
        the duration of the timeout following the restart is set to
        an unspecified value that does not exceed the original
        timeout value.  If the flag is not set, interruptible
        functions interrupted by this signal shall fail with errno
        set to [EINTR].

https://pubs.opengroup.org/onlinepubs/9699919799/functions/sigaction.html

Nothing in the POSIX definition of open specifies otherwise.

In 1990, Kirk McKusick added these lines with a mysterious commit
message:

Author: Kirk McKusick <mckusick>
Date:  Tue Apr 10 19:36:33 1990 -0800

    eliminate longjmp from the kernel (for karels)

diff --git a/sys/kern/vfs_syscalls.c b/sys/kern/vfs_syscalls.c
index 7bc7b39bbf..d572d3a32d 100644
--- a/sys/kern/vfs_syscalls.c
+++ b/sys/kern/vfs_syscalls.c
@@ -14,7 +14,7 @@
  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  *
- * @(#)vfs_syscalls.c 7.42 (Berkeley) 3/26/90
+ * @(#)vfs_syscalls.c 7.43 (Berkeley) 4/10/90
  */

#include "param.h"
@@ -530,8 +530,10 @@ copen(scp, fmode, cmode, ndp, resultfd)
if (error = vn_open(ndp, fmode, (cmode & 07777) &~ S_ISVTX)) {
crfree(fp->f_cred);
fp->f_count--;
- if (error == -1) /* XXX from fdopen */
- return (0); /* XXX from fdopen */
+ if (error == EJUSTRETURN) /* XXX from fdopen */
+ return (0); /* XXX from fdopen */
+ if (error == ERESTART)
+ error = EINTR;
scp->sc_ofile[indx] = NULL;
return (error);
}

(found via this git import of the CSRG history:
https://github.com/robohack/ucb-csrg-bsd/commit/cce2869b7ae5d360921eb411005b328a29c4a3fe)

This change appears to have served two related purposes:

1. The fdopen function (the erstwhile open routine for /dev/fd/N)
  used to return -1 as a hack to mean it had just duplicated the fd;
  it was recently changed by Mike Karels, in kern_descrip.c 7.9, to
  return EJUSTRETURN, now defined to be -2, presumably to avoid a
  conflict with ERESTART, defined to be -1.  So this change finished
  part of the change by Mike Karels to use a different magic return
  code from fdopen.

  Of course, today we use still another disgusting hack, EDUPFD, for
  the same purpose, so none of this is relevant any more.

2. Prior to April 1990, the kernel handled signals during tsleep(9)
  by longjmping out to the system call entry point or similar.  In
  April 1990, Mike Karels worked to convert all of that into
  explicit unwind logic by passing through EINTR or ERESTART as
  appropriate, instead of setjmp at each entry point.

However, it's not clear to me why this setjmp/longjmp and
fdopen/-1/EJUSTRETURN renovation justifies unconditional logic to map
ERESTART to EINTR in open(2).  I suspect it was a mistake.

In 2013, the corresponding logic to map ERESTART to EINTR in open(2)
was removed from FreeBSD:

  r246472 | kib | 2013-02-07 14:53:33 +0000 (Thu, 07 Feb 2013) | 11 lines

  Stop translating the ERESTART error from the open(2) into EINTR.
  Posix requires that open(2) is restartable for SA_RESTART.

  For non-posix objects, in particular, devfs nodes, still disable
  automatic restart of the opens. The open call to a driver could have
  significant side effects for the hardware.

  Noted and reviewed by:  jilles
  Discussed with: bde
  MFC after:      2 weeks

Index: vfs_syscalls.c
===================================================================
--- vfs_syscalls.c (revision 246471)
+++ vfs_syscalls.c (revision 246472)
@@ -1106,8 +1106,6 @@
goto success;
}

- if (error == ERESTART)
- error = EINTR;
goto bad;
}
td->td_dupfd = 0;

https://cgit.freebsd.org/src/commit/sys/kern/vfs_syscalls.c?id=2ca49983425886121b506cb5126b60a705afc38c

It's not clear to me that there's any reason to treat device nodes
specially here; in fact, if a driver's .d_open routine sleeps and is
woken by a concurrent revoke without a signal pending or with an
SA_RESTART signal pending, it is wrong for it to fail with EINTR.
But it MUST restart the whole system call rather than continue
sleeping in a loop or just exit the loop and continue to open,
because it is mandatory in the security model of revoke for open(2)
to retry the permissions check at that point.

PR kern/57260

XXX pullup-8
XXX pullup-9
XXX pullup-10

(riastradh)

Tickets #110 - #112

(martin)

Pull up following revision(s) (requested by mlelstv in ticket #112):

sbin/nvmectl/nvmectl.h: revision 1.10
sbin/nvmectl/logpage.c: revision 1.11
sbin/nvmectl/util.c: revision 1.3

Data units read/written are counted in 1000s of 512 bytes.

Convert to human-readable value.

(martin)

Pull up following revision(s) (requested by hannken in ticket #111):

sys/miscfs/genfs/genfs_vnops.c: revision 1.220

Fix genfs_can_chtimes() to also handle the condition:

  If the time pointer is null, then write permission
  on the file is also sufficient.

From FreeBSD.

Should fix PR kern/57246 "NFS group permissions regression"

(martin)

Pull up following revision(s) (requested by hannken in ticket #110):

external/cddl/osnet/dist/uts/common/fs/zfs/zfs_vnops.c: revision 1.81

Adapt zfs_netbsd_access() to ACL support.  As ZFS itself only
handles VREAD, VWRITE, VEXEC and VAPPEND we use kauth_authorize_vnode()
to handle VADMIN.

From FreeBSD.

(martin)

In the HUP-wait path in ucomopen():
- Use cv_timedwait() rather than cv_timedwait_sig(); the wait here is
  bounded (and fairly short besides) and seems appropriate to treat like
  other uninterruptible waits.  The behavior is now consistent with com(4)
  in this regard.
- Map EWOULDBLOCK return from cv_timedwait() to 0, as the successful passage
  of time is not an error in this case.
- If the HUP-wait time has passed, clear the HUP-wait timestamp.

kern/57259 (although insufficient -- another change to vfs_syscalls.c
is required)

(thorpej)

tests/make: improve explanations in test for 'empty' function

(rillig)

tests/make: use proper variable names in short-circuit test

The previous variable names V42, V66, iV1 and iV2 didn't carry enough
information to be readily readable, making the test hard to understand.

Rename the variables to be more expressive.  While here, properly
explain what happened behind the scenes in 2020 and how the evaluation
of conditions was fixed after discovering the actual cause of the
unexpected error messages.

(rillig)

Trailing whitespace

(skrll)

tests/make: test very small and very large numbers in conditions

(rillig)

Ticket #1609

(martin)

Pull up following revision(s) (requested by sekiya in ticket #1609):

sys/arch/arm/ti/if_cpsw.c: revision 1.17

Uncomment and protect sanity checks that would drop into the debugger with a
CPSW_DEBUG_DMA define.

This handles a condition where checking for DMA_RXEOQ in the received packet
results in the console being spammed with "rxeoq" messages, which soon results
in a kernel panic.  The corresponding Debugger() call for this check was
commented out.

The TI documentation ("AM335x and AMIC110 Sitara(TM) Processors Technical
Reference Manual") documents the EOQ bit thus:

        (14.3.2.4.1.2.4) This bit is set by the EMAC when the EMAC identifies
        that a descriptor is the last for a given packet received (also sets
        the EOP flag), and there are no more descriptors in the receive list
        (next descriptor pointer is NULL).  The software application can use
        this bit to detect when the EMAC receiver for the corresponding channel
        has halted. This is useful when the application appends additional free
        buffer descriptors to an active receive queue. Note that this flag is
        valid on EOP descriptors only.

Moving this check (and the offending printf() ) into a debug #ifdef results in
the BeagleBone Green surviving the nightly checks; it would always drop into
the debugger before this change.

(martin)

Ticket #109

(martin)

Pull up following revision(s) (requested by sekiya in ticket #109):

sys/arch/arm/ti/if_cpsw.c: revision 1.17

Uncomment and protect sanity checks that would drop into the debugger with a
CPSW_DEBUG_DMA define.

This handles a condition where checking for DMA_RXEOQ in the received packet
results in the console being spammed with "rxeoq" messages, which soon results
in a kernel panic.  The corresponding Debugger() call for this check was
commented out.

The TI documentation ("AM335x and AMIC110 Sitara(TM) Processors Technical
Reference Manual") documents the EOQ bit thus:

        (14.3.2.4.1.2.4) This bit is set by the EMAC when the EMAC identifies
        that a descriptor is the last for a given packet received (also sets
        the EOP flag), and there are no more descriptors in the receive list
        (next descriptor pointer is NULL).  The software application can use
        this bit to detect when the EMAC receiver for the corresponding channel
        has halted. This is useful when the application appends additional free
        buffer descriptors to an active receive queue. Note that this flag is
        valid on EOP descriptors only.

Moving this check (and the offending printf() ) into a debug #ifdef results in
the BeagleBone Green surviving the nightly checks; it would always drop into
the debugger before this change.

(martin)

add CMC (Certificate Management over CMS, RFC5272)
add CMS (Cryptographic Message Syntax, RFC5652)

(jschauma)

x86/fpu: Align savefpu to 64 bytes in fpuinit_mxcsr_mask.

16 bytes is not enough.

(Is this why it never worked on Xen some years back?  Got lucky and
accidentally had 64-byte alignment on native x86, but not in the call
stack in Xen?)

XXX pullup-10

(riastradh)

x86: Call fpuinit_mxcsr_mask only once.

No need to call it again and again on the secondary CPUs to compute
what should be the same mxcsr mask.  (If it's not, we have deeper
problems!)

(riastradh)

Revert "x86: Add kthread_fpu_enter/exit support, take two."

kthread_fpu_enter/exit changes broke some hardware, unclear why, to
investigate before fixing and reapplying these changes.

(riastradh)

Revert "x86/fpu.c: Sprinkle KNF."

kthread_fpu_enter/exit changes broke some hardware, unclear why, to
investigate before fixing and reapplying these changes.

(riastradh)

Revert "x86: Add kthread_fpu_enter/exit support, take two -- forgot i386 bits."

kthread_fpu_enter/exit changes broke some hardware, unclear why, to
investigate before fixing and reapplying these changes.

(riastradh)

mremap(2): Note MAP_REMAPDUP bug.

XXX pullup-10, unless we fix the bug first

(riastradh)

entropy(9): Allow changing flags on all entropy sources at once.

Entropy sources should all have nonempty names, and this will enable
an operator to, for example, disable all but a specific entropy
source.

XXX pullup-10

(riastradh)

Fix genfs_can_chtimes() to also handle the condition:

  If the time pointer is null, then write permission
  on the file is also sufficient.

From FreeBSD.

Should fix PR kern/57246 "NFS group permissions regression"

(hannken)

Adapt zfs_netbsd_access() to ACL support.  As ZFS itself only
handles VREAD, VWRITE, VEXEC and VAPPEND we use kauth_authorize_vnode()
to handle VADMIN.

From FreeBSD.

(hannken)

fix idempotent check define as noticed by LLVM.

(mrg)

delete attribution per request of contributor

(nia)

delete attribution per request of contributor

(nia)

delete attribution per request of contributor

(nia)

delete attribution per request of contributor

(nia)

mremap(2): Fix example to use MAP_PRIVATE.

It is a historical accident that MAP_PRIVATE is assumed when neither
it nor MAP_SHARED is specified.

XXX pullup-9
XXX pullup-10

(riastradh)

fs/hfs: Avoid undefined pointer arith in hfslib_reada_node_offsets.

XXX pullup-8
XXX pullup-9
XXX pullup-10

(riastradh)

fs/hfs: Avoid buffer overrun in hfslib_reada_node_offsets.

XXX pullup-8
XXX pullup-9
XXX pullup-10

(riastradh)

Add RTTI (run-time type information).

(fcambus)

Fix a typo in a newly added comment.

Someone please tell me why these things become obvious only after
the code has been committed!

NFC

(kre)

When processing swapon -a (or swapctl -A, or swapctl -U) ignore lines in
fstab that have nothing to do with swapping (fs_type is neither "sw" nor "dp")
before running getfsspecname() on the fs_spec field of the line.

This avoids entries like this:

NAME=OFTEN_UNCONNECTED  /local/archived ffs    rw,log,noauto    0 0

in fstab from generating spurious error messages when the wedge named
is not currently connected to the system - that is the drive on which the
wedge exists is not connected, or not powered on.  "noauto" handles that
for some other uses, the "0"s in fs_freq and fs_passno work for other uses,
but swap{on,ctl} never look at those fields (not for this purpose).

Non "sw"/"dp" lines were being ignored anyway, but not until (a little) later.

(kre)

mmap(2): minor markup tweaks

(uwe)

Add TU (translation unit).

(fcambus)

Add RAII (resource acquisition is initialization).

(fcambus)

Add IR (intermediate representation) and MLIR (multi-level IR).

(fcambus)

nouveau: Kick out genfb on firmware framebuffer before initializing.

PR kern/53126

(riastradh)

x86: Expand on comments on ordering around stores to ci_curlwp.

No functional change intended.

PR kern/57240

(riastradh)

sparc64: Optimization: Omit needless membar when triggering softint.

When we are triggering a softint, it can't already hold any mutexes.
So any path to mutex_exit(mtx) must go via mutex_enter(mtx), which is
always done with atomic r/m/w, and we need not issue any explicit
barrier between ci->ci_curlwp = softlwp and a potential load of
mtx->mtx_owner in mutex_exit.

PR kern/57240

XXX pullup-8
XXX pullup-9
XXX pullup-10

(riastradh)

riscv: Optimization: Omit needless membar when triggering softint.

When we are triggering a softint, it can't already hold any mutexes.
So any path to mutex_exit(mtx) must go via mutex_enter(mtx), which is
always done with atomic r/m/w, and we need not issue any explicit
barrier between ci->ci_curlwp = softlwp and a potential load of
mtx->mtx_owner in mutex_exit.

PR kern/57240

(riastradh)

powerpc: Optimization: Omit needless membar when triggering softint.

When we are triggering a softint, it can't already hold any mutexes.
So any path to mutex_exit(mtx) must go via mutex_enter(mtx), which is
always done with atomic r/m/w, and we need not issue any explicit
barrier between ci->ci_curlwp = softlwp and a potential load of
mtx->mtx_owner in mutex_exit.

PR kern/57240

XXX pullup-8
XXX pullup-9
XXX pullup-10

(riastradh)

mips: Optimization: Omit needless membar when triggering softint.

When we are triggering a softint, it can't already hold any mutexes.
So any path to mutex_exit(mtx) must go via mutex_enter(mtx), which is
always done with atomic r/m/w, and we need not issue any explicit
barrier between ci->ci_curlwp = softlwp and a potential load of
mtx->mtx_owner in mutex_exit.

PR kern/57240

XXX pullup-8
XXX pullup-9
XXX pullup-10

(riastradh)

arm32: Optimization: Omit needless membar when triggering softint.

When we are triggering a softint, it can't already hold any mutexes.
So any path to mutex_exit(mtx) must go via mutex_enter(mtx), which is
always done with atomic r/m/w, and we need not issue any explicit
barrier between ci->ci_curlwp = softlwp and a potential load of
mtx->mtx_owner in mutex_exit.

PR kern/57240

XXX pullup-8
XXX pullup-9
XXX pullup-10

(riastradh)

aarch64: Optimization: Omit needless membar when triggering softint.

When we are triggering a softint, it can't already hold any mutexes.
So any path to mutex_exit(mtx) must go via mutex_enter(mtx), which is
always done with atomic r/m/w, and we need not issue any explicit
barrier between ci->ci_curlwp = softlwp and a potential load of
mtx->mtx_owner in mutex_exit.

PR kern/57240

XXX pullup-9
XXX pullup-10

(riastradh)