vm_page: alloc_domain_iter
A recent change that made a call to pctrie_iter_reset conditional was
ill-advised, as it leads to an assertion failure on the graid1_6.sh
stress2 test. Make it unconditional again.
Reported by: pho
Fixes: 3fe2f81f665b ("vm_page: reduce iter_reset calls in alloc_domain_iter")
seq.1: Reset option list alignment + tag spdx
This option list was tabbed over 19 characters. style.mdoc(5) asks
that lists are set to the longest item, or if that's too long then
to use indent. Resetting it to indent buys us two lines at MANWIDTH
80, and three at 59.
MFC after: 3 days
Approved by: mhorne (mentor)
Differential Revision: https://reviews.freebsd.org/D48350
(cherry picked from commit 2f893597fcc34456d5adfc3f1633b39c76d0764f)
style: Recomend SPDX licenses
While here, take this opportunity to update the copyright and clean
some mdoc errors.
MFC after: 3 days
Reviewed by: carlavilla, mhorne, imp
Approved by: imp (srcmgr)
Approved by: carlavilla, mhorne (mentors)
Differential Revision: https://reviews.freebsd.org/D49051
(cherry picked from commit b3de609802104234dca7449fc2b4f4106098d41b)
UPDATING: Strengthen recommendations
Strengthen recommendations for ports_modules by adding a separate note.
Use the unused [2] to add a note about creating backups.
MFC after: 3 days
Reviewed by: brooks, mhorne
Approved by: mhorne (mentor)
Differential Revision: https://reviews.freebsd.org/D50126
(cherry picked from commit ca9131b729757959cb0fd4721603466284cfc6fe)
e6060sw.4: Improve title and add HARDWARE
For visibility in apropos and inclusion in the hardware release notes.
MFC after: 3 days
Reviewed by: adrian, mhorne
Approved by: mhorne (mentor)
Differential Revision: https://reviews.freebsd.org/D50321
(cherry picked from commit bef89c155e77dd471af596a43f3914b754cc5155)
ar40xx.4: Initial manual page
For inclusion in apropos and the hardware release notes. While here,
add xrefs to this page and the previous e6000sw in etherswitch.4.
MFC after: 3 days (but adjusted for 14.3)
Reported by: adrian
Reviewed by: adrian, mhorne
Approved by: mhorne (mentor)
Differential Revision: https://reviews.freebsd.org/D50324
(cherry picked from commit 91c9751c3913497db2fbad1fcf7491557bbd2a75)
arswitch.4: Improve title and add HARDWARE
For visibility in apropos and inclusion in the hardware release notes.
MFC after: 3 days
Reviewed by: adrian (previous), mhorne
Approved by: mhorne
(cherry picked from commit 9692d5e15e9bd7e628b06d655522477ce8c5b6fe)
qat.4: Cleanup
+ copyright symbols do not belong in Nd's (apropos results)
+ move hardware to HARDWARE for inclusion in the hw release notes
+ improve sysctls
+ lines which say nothing removed
+ links moved to see also
+ spacing and markup corrections
MFC after: 3 days
Reviewed by: markj, mhorne
Differential Revision: https://reviews.freebsd.org/D50381
grantbylabel_syscall check p_textvp != NULL
kernel process will not have valid p_textvp
Reviewed by: stevek
Differential Revision: https://reviews.freebsd.org/D50368
nfs_clrpcops.c: Fix handling of a CreateLayGet reply
Without this patch, the CreateLayGet RPC did not handle a NFSv4.1/4.2
reply for a pNFS server that also issued delegations.
This patch fixes the problem.
Detected at the recent NFSv4 Bakeathon testing event.
This bug would only affect the rare case where the FreeBSD client
is mounted to a pNFS server that issues delegations, where the
"pnfs" mount option is specified.
MFC after: 2 weeks
traceroute tests: fix the test when SCTP is supported
The SCTP test assumes a default system configuration where SCTP is not
supported, so the probe packet returns an ICMP error which is displayed
as a !P response. If SCTP is supported, then something else is returned
instead (depending on exactly what probe we sent, but not an ICMP error)
and the test fails.
Since we already check the correct probes are sent using tcpdump, remove
the match for the second hop entirely.
Reported by: Jenkins
Reviewed by: des
Approved by: des (mentor)
Differential Revision: https://reviews.freebsd.org/D50336
man: Exit cleanly on SIGPIPE.
The first attempt at addressing this simply suppressed SIGPIPE, which
resulted in mandoc printing out error messages instead. This was then
reverted, but the pipefail was (correctly) left in, so man still
returned a nonzero exit code if you quit a page before the end.
PR: 223516, 279542
Fixes: 14a5c1068d37, a85d870007e7
MFC after: 1 week
Reviewed by: ziaee, kevans
Differential Revision: https://reviews.freebsd.org/D50302
sh: Avoid referencing uninitialized memory in alias
If run as
alias ''
uninitialized memory could be referenced.
This is based on a fix from NetBSD. For more information, refer to
https://github.com/NetBSD/src/commit/10cfed82c28 .
Obtained from: NetBSD (Robert Elz <kre at netbsd.org>, 10cfed82c28)
MFC after: 3 days
Reported by: mckusick, Robert Elz <kre at netbsd.org>
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D50364
ppt: Fix panic when configuring unavailable MSI-X vector
In some cases VM may have different idea about number
of available MSI-X vectors then PPT driver. Return
an error when VM requests setup for more vectors
than expected.
It was observed while using SR-IOV on an Intel E810 Ethernet adapter.
VF driver in a VM sees a correct number of available MSI-X vectors,
which depends on num-queues assigned in iovctl.conf, while
pci_msix_count in the PPT driver always returns 1.
Signed-off-by: Krzysztof Galazka <kgalazka at FreeBSD.org>
Reviewed By: #bhyve, jhb
Approved by: erj (mentor), kbowling (mentor)
MFC after: 1 week
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D48812
stress2: Fix scenario after the changes to ptrace(PT_ATTACH) in commit ecc662c749b1.
The traced process now no longer exits immediately at the attach.
After a PT_DETACH do not try to wait on a non child process. Style fixes.
vm_page: conditionally compile label
A change that removed a goto left a label the target of no gotos on powerpc.
Use the preprocessor to hide that label when it's not a target.
Fixes: 7c9fe7a7925b ("vm_page: remove obsolete page alloc loops")
vm_page: reduce iter_reset calls in alloc_domain_iter
Where the iterator is reset in all cases now, it actually only needs
to be reset when ALLOC_WAITFAIL is set.
Reported by: alc
Reviewed by: alc, kib, markj
Differential Revision: https://reviews.freebsd.org/D50263
vm_page: use vm_page_alloc_fail consistently
The function vm_page_alloc_fail is used in different ways in different
places to achieve the same effect. This purely stylistic change makes
the usages consistent.
Reviewed by: alc, kib, markj
Differential Revision: https://reviews.freebsd.org/D50263