Let firmware do its best first, and if it can't, try software recovery.
I would remove software timeout handler completely, but found bunch of
complains on command timeout on sparc64 mailing list few years ago, so
better be safe in case of interrupt loss.
MFC r315579, r315670: Add initial support for multiple MSI-X vectors.
For 24xx and above use 2 vectors (default and response queue).
For 26xx and above use 3 vectors (default, response and ATIO queues).
Due to global lock interrupt hardlers never run simultaneously now, but
at least this allows to save one regitster read per interrupt.
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
* In zfs_freebsd_setattr, if the caller wants to set the birthtime,
set the bits that zfs_settattr expects
* In zfs_setattr, if XAT_CREATETIME is set, set xoa_createtime,
expected by zfs_xvattr_set. The two levels of indirection seem
excessive, but it minimizes diffs vs OpenZFS.
* In zfs_setattr, check for overflow of va_birthtime (from delphij)
* Remove red herring in zfs_getattr
sys/cddl/contrib/opensolaris/uts/common/sys/vnode.h
* Un-booby-trap some macros
New tests are under review at https://github.com/pjd/pjdfstest/pull/6
ATF tests have a default WARNS of 0, unlike other usermode programs. This
change is technically a noop, but it documents that the msun tests don't
work with any warnings enabled, at least not on all architectures.
r310146:
Use the right bitwise OR operation for clearing single-step at trap time.
r311912:
Force all TOC references in asm to include '@toc'
r312369:
Use the explicit expanded form of cmp.
r312617:
Hide the 'MOREARGS' macro, it conflicts with contrib code, and is only used
in one file.
r312614:
Don't pass -Wa,-many through clang, the integrated as doesn't support it.
r312659:
Avoid using non-zero argument for __builtin_frame_address().
r312974:
Add a INTR_TRIG_INVALID, and use it in the powerpc interrupt code.
r312977:
Force the setting of bit 7 in the sysmouse packet byte 1 to be unsigned.
r313005:
Update CFLAGS for clang compatibility
r314826:
Clang in base now supports -mlongcall, so remove this hack
Open Firmware runs in virtual mode on the Powermac G5. This runs inside
the
kernel page table, which preserves all address translations made by OF
before
the kernel starts; as a result, the kernel address space is a strict
superset of
OF's.
Where this explodes is if OF uses an unmapped SLB entry. The SLB
fault handler
runs in real mode and refers to the PCPU pointer in SPRG0,
which blows up the
kernel. Having a value of SPRG0 that works for the kernel is
less fatal than
preserving OF's value in this case.
===
The result of this is seemingly random panics from
NULL dereferences, or hangs
immediately upon boot. By not restoring SPRG0 for
Open Firmware entry the
kernel PCPU pointer is preserved and SLB faults
are successful, resulting in a
stable kernel.
Change the "devfs_fsync: vop_stdfsync failed" from panic to a printf.
It's not a proper fix, but should be better than what we have now.
Since it got broken some six months ago it results in an incredibly
annoying and trivially reproducible panic every time eg an USB disk
gets disconnected.
Get rid of foo_sys() in linuxulator code. It was commented out, and it
would be useless anyway - there is no point in pretending to have block
devices; our "block" devices are in fact character ones, and can only
be accessed as such.
Most of these are null pointer dereferences or missing error checks in the
unit tests. One is a missing error check in xnb_attach_failed. None can
cause real problems in running systems.
If one of the scripts listed in (daily|weekly|monthly)_local is executable,
999.local should simply execute it. Only if the script isn't executable
should 999.local assume it needs /bin/sh.
Interesting fixes which were not already merged: 0c7c611 Merge C++ demangler bug fixes from ELF Tool Chain (#40) 2b208d9 __cxa_demangle_gnu3: demangle 'z' as '...', not 'ellipsis' (#41)
mm [Fri, 31 Mar 2017 20:16:24 +0000 (20:16 +0000)]
MFC r315636,315876,316095:
Sync libarchive with vendor
Vendor changes/bugfixes (FreeBSD-related):
r315636:
PR 867 (bsdcpio): show numeric uid/gid when names are not found
PR 870 (seekable zip): accept files with valid ZIP64 EOCD headers
PR 880 (pax): Fix handling of "size" pax header keyword
PR 887 (crypto): Discard 3072 bytes instead of 1024 of first keystream
OSS-Fuzz issue 806 (mtree): rework mtree_atol10 integer parser
Break ACL read/write code into platform-specific source files
r315876:
Store extended attributes with extattr_set_link() if no fd is provided
Add extended attribute tests to libarchive and bsdtar
Fix tar's test_option_acls
Support the UF_HIDDEN file flag
r316095:
Constify variables in several places
Unify platform ACL code in a single source file
Fix unused variable if compiling on FreeBSD without NFSv4 ACL support
ed [Fri, 31 Mar 2017 08:43:07 +0000 (08:43 +0000)]
MFC r315892:
Include <sys/systm.h> to obtain the memcpy() prototype.
I got a report of this source file not building on Raspberry Pi. It's
interesting that this only fails for that target and not for others.
Again, that's no reason not to include the right headers.
truckman [Fri, 31 Mar 2017 06:20:06 +0000 (06:20 +0000)]
MFC r315516
Change several constants used by the PIE algorithm from unsigned to signed.
- PIE_MAX_PROB is compared to variable of int64_t and the type promotion
rules can cause the value of that variable to be treated as unsigned.
If the value is actually negative, then the result of the comparsion
is incorrect, causing the algorithm to perform poorly in some
situations. Changing the constant to be signed cause the comparision
to work correctly.
- PIE_SCALE is also compared to signed values. Fortunately they are
also compared to zero and negative values are discarded so this is
more of a cosmetic fix.
- PIE_DQ_THRESHOLD is only compared to unsigned values, but it is small
enough that the automatic promotion to unsigned is harmless.
dchagin [Thu, 30 Mar 2017 20:12:23 +0000 (20:12 +0000)]
MFC r314402:
FreeBSD does not have analgue for epoll EPOLLPRI event type.
So, do not set EPOLLPRI event acidently.
Also, do not set EPOLLWRNORM and EPOLLRDNORM events as epoll
do not set this events.
dchagin [Thu, 30 Mar 2017 20:00:57 +0000 (20:00 +0000)]
MFC r314293:
Return EOVERFLOW error in case then the size of tv_sec field of struct timespec
in COMPAT_LINUX32 Linuxulator's not equal to the size of native tv_sec.
tsoome [Thu, 30 Mar 2017 17:23:40 +0000 (17:23 +0000)]
boot1.efi: can't boot from ZFS on 4kn HDD
The boot1.efi immediate issue from PR216964 is that we are reading into
too small buffer, from UEFI spec 2.6:
The size of the Buffer in bytes. This must be a multiple of the intrinsic block size of the device.
The secondary issue is that LBA calculation does not check reminder from
division.
This fix does check the provided buffer size and if we read less than
media sector size or the read offset is not aligned to sector boundary,
we allocate bounce buffer and perform the read by single sector.
ae [Thu, 30 Mar 2017 14:20:27 +0000 (14:20 +0000)]
MFC r303018:
Add named dynamic states support to ipfw(4).
The keep-state, limit and check-state now will have additional argument
flowname. This flowname will be assigned to dynamic rule by keep-state
or limit opcode. And then can be matched by check-state opcode or
O_PROBE_STATE internal opcode. To reduce possible breakage and to maximize
compatibility with old rulesets default flowname introduced.
It will be assigned to the rules when user has omitted state name in
keep-state and check-state opcodes. Also if name is ambiguous (can be
evaluated as rule opcode) it will be replaced to default.
MFC r304087:
Do not warn about ambiguous state name when we inspect a comment token.
MFC r304089:
Add an ability to attach comment to check-state rules.
MFC r310727 (by marius):
Fix a bug in r272840; given that the optlen parameter of setsockopt(2)
is a 32-bit socklen_t, do_get3() passes the kernel to access the wrong
32-bit half on big-endian LP64 machines when simply casting the 64-bit
size_t optlen to a socklen_t pointer.
While at it and given that the intention of do_get3() apparently is to
hide/wrap the fact that socket options are used for communication with
ipfw(4), change the optlen parameter of do_set3() to be of type size_t
and as such more appropriate than uintptr_t, too.
MFC r315305:
Change the syntax of ipfw's named states.
Since the state name is an optional argument, it often can conflict
with other options. To avoid ambiguity now the state name must be
prefixed with a colon.
r314547
loader.efi: reduce the size of the staging area if necessary
The loader assumes physical memory in [2MB, 2MB + EFI_STAGING_SIZE)
is Conventional Memory, but actually it may not, e.g. in the case
of Hyper-V Generation-2 VM (i.e. UEFI VM) running on Windows
Server 2012 R2 host, there is a BootServiceData memory block at
the address 47.449MB and the memory is not writable.
Without the patch, the loader will crash in efi_copy_finish():
see PR 211746.
The patch verifies the end of the staging area, and reduces its
size if necessary. This way, the loader will not try to write into
the BootServiceData memory any longer.
Thank Marcel Moolenaar for helping me on this issue!
The patch also allocates the staging area in the first 1GB memory.
See the comment in the patch for this.
r314770
loader.efi: fix recent UEFI-boot regression on physical machines
This patch fixes my recent patch
"loader.efi: reduce the size of the staging area if necessary", which
causes EFI-boot failure on physical machines since Mar 2:
on the host there is a 1MB LoaderData memory range, which splits
the big Conventional Memory range into a small one (15MB) and a
big one: the small one is too small to hold the staging area.
We can actually use the LoaderData range safely, because when
amd64_tramp -> efi_copy_finish() starts to run, we're almost at
the very end of the efi loader code and we're going to "return"
to the kernel entry, so we're pretty sure we won't access any loader
data any more.
For people who are interested in the details: please see
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=211746#c22
PS, some people also reported the regression happened to FreeBSD VM
running on Bhyve in EFI mode. This patch should resolve it too,
though I don't have such a setup to test.
r314828
loader.efi: fix an off-by-one bug in efi_verify_staging_size()
Also remove the warning message: it may not be unusual to see
the memory range containing 2MB is not of EfiConventionalMemory.
Sponsored by: Microsoft
r314891
loader.efi: finally fix the off-by-one bug in efi_verify_staging_size()
r314828(loader.efi: fix an off-by-one bug in efi_verify_staging_size())
doesn't really fix the bug and this patch adds the missing part.
It's a shame that I didn't make everything correct at the very beginning...
Sponsored by: Microsoft
r314956
loader.efi: only reduce the size of the staging area on Hyper-V
Doing this on physical hosts turns out to be problematic, e.g. see comment
24 and 28 in https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=211746.
To fix the real underlying issue correctly & thoroughly, IMO we need
a relocatable kernel, but that would require a lot of complicated long
term work: https://reviews.freebsd.org/D9686?id=25414#inline-56969
For now, let's only apply efi_verify_staging_size() to VMs running on
Hyper-V, and restore the old behavior on physical machines since that
has been working for people for a long period of time, though that's
potentially unsafe...
Sponsored by: Microsoft
r314962
loader.efi: only include the machine/ header files on x86
The 2 files may not exist on other archs like aarch64 and hence we
can have a build failure there.
Reported by: lwhsu
Sponsored by: Microsoft
r315235
loader.efi: use stricter check for Hyper-V
Some other hypervisors like Xen can pretend to be Hyper-V but obviously
they can't implement all Hyper-V features. Let's make sure we're genuine
Hyper-V here.
ngie [Thu, 30 Mar 2017 05:47:18 +0000 (05:47 +0000)]
Backport mlx4{en,ib}(4) from ^/head
MFCing other pieces would be very structurally disruptive. This just
brings back the manpages so they can be used by end-users and to ease
future backports.
svn:mergeinfo omitted, in part because this is a direct commit to
^/stable/11.
ngie [Thu, 30 Mar 2017 05:14:52 +0000 (05:14 +0000)]
MFC r314372:
Use "build" instead of "all" when building ports modules
"all" in ports currently means "stage the ports", which requires root today,
and brings to light other potential issues, like ENAMETOOLONG with staged
directories (bug 161481, etc).
This fixes buildkernel for me when run as a non-root user, assuming all
of the prerequisites have been installed beforehand and are up-to-date.
gnn [Thu, 30 Mar 2017 02:50:21 +0000 (02:50 +0000)]
MFC: 313176, 313177, 313359
Replace the implementation of DTrace's RAND subroutine for generating
low-quality random numbers with a modern implementation (xoroshiro128+)
that is capable of generating better quality randomness without compromising performance.
gnn [Thu, 30 Mar 2017 02:38:30 +0000 (02:38 +0000)]
MFC: 311225, 311243, 313045
Fix DTrace TCP tracepoints to not use mtod() as it is both unnecessary and
dangerous. Those wanting data from an mbuf should use DTrace itself to get
the data.
Add an mbuf to ipinfo_t translator to finish cleanup of mbuf passing to TCP probes.
projects / FreeBSD / FreeBSD.git / log