Kyle Evans [Sun, 31 Jan 2021 15:51:39 +0000 (09:51 -0600)]
lualoader: position hyphens at the beginning of character classes
According to the Lua 5.4 manual section 6.4.1 ("Patterns"), the interaction
between ranges and classes is not defined and hyphens must be specified at
either the beginning or the end of a set if they are not escaped.
Kyle Evans [Sun, 24 Jan 2021 01:32:38 +0000 (19:32 -0600)]
stand: lua: enhance lfs.dir() to speed up kernels_autodetect
This eliminates a lot of stat() calls that happen when lualoader renders the
menu with the default settings, and greatly speeds up rendering on my
laptop.
ftype is nil if loader/loader.efi hasn't been updated yet, falling back to
lfs.attributes() to test.
This is technically incompatible with lfs, but not in a particularly
terrible way.
MFC: mips: fix early kernel panic when setting up interrupt counters
Commit 248f0ca converted intrcnt and intrnames from u_long[]
and char[] to u_long* and char* respectively, but for non-INTRNG mips
these symbols were defined in .S file as a pre-allocated static arrays,
so the problem wasn't cought at compile time. Conversion from an array
to a pointer requires pointer initialization and it wasn't done
for MIPS, so whatever happenned to be in the begginning of intcnt[]
array was used as a pointer value.
Move intrcnt/intrnames to C code and allocate them dynamically
although with a fixed size at the moment.
Alexander Motin [Sun, 31 Jan 2021 17:46:57 +0000 (12:46 -0500)]
cxgb(4): Remove assumption of physically contiguous mbufs.
Investigation of iSCSI target data corruption reports brought me to
discovery that cxgb(4) expects mbufs to be physically contiguous, that
is not true after I've started using m_extaddref() in software iSCSI
for large zero-copy transmissions. In case of fragmented memory the
driver transmitted garbage from pages following the first one due to
simple use of pmap_kextract() for the first pointer instead of proper
bus_dmamap_load_mbuf_sg(). Seems like it was done as some optimization
many years ago, and at very least it is wrong in a world of IOMMUs.
This patch just removes that optimization, plus limits packet coalescing
for mbufs crossing page boundary, also depending on assumption of one
segment per packet.
Toomas Soome [Sun, 24 Jan 2021 18:59:36 +0000 (20:59 +0200)]
vt: panic while changing vt font
Set refcount for loader provided font to 1 to prevent this font
from being released (so we can reset to default).
As we get started with no memory allocator, we set up static font data
for font passed by loader (if there is any). At this time, we also must
set refcount 1, and refcount will get incremented in cnprobe() callback.
At some point the memory allocator will be available, and we will set up
properly allocated font data, but we should not disturb the refcount.
Marcin Wojtas [Tue, 26 Jan 2021 10:49:35 +0000 (11:49 +0100)]
marvell: ap806_clock: add missing frequency modes
In the driver init routine the CPU clock frequency
value is obtained from a dedicated register. Until now
only part of the values were handled by the mv_ap806_clock
driver. Fix that by adding missing cases.
Bjoern A. Zeeb [Thu, 28 Jan 2021 16:32:43 +0000 (16:32 +0000)]
LinuxKPI: implement devres() framework parts and two examples
This code implements a version of the devres framework found
working for various iwlwifi use cases and also providing functions
for ttm_page_alloc_dma.c from DRM.
Part of the framework replicates the consumed KPI, while others
are internal helper functions.
In addition the simple devm_k*malloc() consumers were implemented
and kvasprintf() was enhanced to also work for the devm_kasprintf()
case.
Addmittingly lkpi_devm_kmalloc_release() could be avoided but for
the overall understanding of the code and possible memory tracing
it may still be helpful.
Further devsres consumer are implemented for iwlwifi but will follow
later as the main reason for this change is to sort out overlap with
DRM.
Sponsored-by: The FreeBSD Foundation
Obtained-from: bz_iwlwifi
MFC After: 3 days
Reviewed-by: hselasky, manu
Differential Revision: https://reviews.freebsd.org/D28189
Bjoern A. Zeeb [Thu, 28 Jan 2021 16:23:19 +0000 (16:23 +0000)]
LinuxKPI: enhance PCI bits for DRM
In pci_domain_nr() directly return the domain which got set in
lkpifill_pci_dev() in all cases. This was missed between D27550
and 105a37cac76b971f7a94409fbdc4f508a7e97fa0 .
In order to implement pci_dev_put() harmonize further code
(which was started in the aforementioned commit) and add kobj
related bits (through the now common lkpifill_pci_dev() code)
to the DRM specific calls without adding the DRM allocated
pci devices to the pci_devices list.
Add a release for the lkpinew_pci_dev() (DRM) case so freeing
will work.
This allows the DRM created devices to use the normal kobj/refcount
logic and work with, e.g., pci_dev_put().
(For a slightly more detailed code walk see the review).
Sponsored-by: The FreeBSD Foundation
Obtained-from: bz_iwlwifi (partially)
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D28188
Bjoern A. Zeeb [Thu, 28 Jan 2021 16:15:12 +0000 (16:15 +0000)]
LinuxKPI: upstream a collection of drm-kmod conflicting changes
The upcoming in-kernel implementations for LinuxKPI based on work on
iwlwifi (and other wireless drivers) conflicts in a few places with
the drm-kmod graphics work outside the base system.
In order to transition smoothly extract the conflicting bits.
This included "unaligned" accessor functions, sg_pcopy_from_buffer(),
IS_*() macros (to be further restricted in the future), power management
bits (possibly no longer conflicting with DRM), and other minor changes.
Obtained-from: bz_iwlwifi Sponsored-by: The FreeBSD Foundation
MFC after: 3 days
Reviewed by: kib, hselasky, manu, bdragon (looked at earlier versions)
Differential Revision: https://reviews.freebsd.org/D26598
Bjoern A. Zeeb [Thu, 28 Jan 2021 16:05:32 +0000 (16:05 +0000)]
LinuxKPI: add firmware loading support
Implement linux firmware KPI compat code.
This includes: request_firmware() request_firmware_nowait(),
request_firmware_direct(), firmware_request_nowarn(),
and release_firmware().
Given we will try to map requested names from natively ported
or full-linuxkpi-using drivers to a firmware(9) auto-loading
name format (.ko file name and image name matching),
we quieten firmware(9) and print success or failure (unless
the _nowarn() version was called) in the linuxkpi implementation.
At the moment we try up-to 4 different naming combinations,
with path stripped, original name, and requested name with '/'
or '.' replaced.
We do not currently defer loading in the "nowait" case.
Sponsored-by: The FreeBSD Foundation
Sponsored-by: Rubicon Communications, LLC ("Netgate")
(firmware(9) nowarn update from D27413)
MFC after: 3 days
Reviewed by: kib, manu (looked at older versions)
Differential Revision: https://reviews.freebsd.org/D27414
Warner Losh [Mon, 25 Jan 2021 19:53:31 +0000 (12:53 -0700)]
newvers: tweak uname to be more useful
The current uname is branch-cXXXX-gHASH
Three changes to make uname more useful.
1. Move from using git rev-list --count to git rev-lis --count --first-parent
since that gives a better, incrementing number.
2. Report this count as 'nXXXXX' rather than 'cXXXXX' because c is part of
a hash and we've changed the sematnics of XXXXX
3. Remove g to make HASH cut and pastable.
Durting review, #1 & #3 had the largest consensus. There was a diversity of
opinion on #2, but on the whole it was positive so I'll acknowledge the dissent,
but move forward with something seems to have support since the dissent was all
about what letter to use where I chose 'n'.
Mitchell Horne [Fri, 15 Jan 2021 16:14:27 +0000 (12:14 -0400)]
bsdinstall: create /efi/boot directory in ESP
If the installer is creating a new ESP, then this directory will not
exist and the subsequent cp will fail silently. This is usually of no
consequence if /efi/freebsd/loader.efi is set up correctly.
diff: fix incorrectly displaying files as duplicates
When diff hits certain access errors, function diffreg() shows the error
message, and then returns to the calling function, which calls
print_status() with the return value.
However, in these cases, the return value isn't changed from the initial
default value of D_SAME.
Normally, print_status() with a value of D_SAME does nothing, so this
works out ok, however, if the "-s" flag is set, a message is displayed
showing identicality:
case D_SAME:
if (sflag)
printf("Files %s%s and %s%s are identical\n", path1, entry, path2, entry);
break;
This then produces such results as:
% diff -s /COPYRIGHT /var/run/rpcbind.sock
diff: /var/run/rpcbind.sock: Operation not supported
Files /COPYRIGHT and /var/run/rpcbind.sock are identical
% diff -s /COPYRIGHT /etc/master.passwd
diff: /etc/master.passwd: Permission denied
Files /COPYRIGHT and /etc/master.passwd are identical
Create a D_ERROR status which is returned in such cases, and
print_status() then deals with that status seperately from D_SAME
Mateusz Guzik [Fri, 29 Jan 2021 21:48:11 +0000 (22:48 +0100)]
Reimplement strlen
The previous code neglected to use primitives which can find the end
of the string without having to branch on every character.
While here augment the somewhat misleading commentary -- strlen as
implemented here leaves performance on the table, especially so for
userspace. Every arch should get a dedicated variant instead.
In the meantime this commit lessens the problem.
Tested with glibc test suite.
Naive test just calling strlen in a loop on Haswell (ops/s):
Mateusz Guzik [Thu, 28 Jan 2021 23:52:08 +0000 (00:52 +0100)]
poll: use fget_unlocked or fget_only_user when feasible
This follows select by eleminating the use of filedesc lock.
This is a win for single-threaded processes and a mixed bag for others
as at small concurrency it is faster to take the lock instead of
refing/unrefing each file descriptor.
Nonetheless, removal of shared lock usage is a step towards a
mtx-protected fd table.
Mateusz Guzik [Tue, 26 Jan 2021 16:19:12 +0000 (17:19 +0100)]
cache: tidy up handling of foo/bar lookups where foo is not a directory
The code was performing an avoidable check for doomed state to account
for foo being a VDIR but turning VBAD. Now that dooming puts a vnode
in a permanent "modify" state this is no longer necessary as the final
status check will catch it.
Mateusz Guzik [Mon, 25 Jan 2021 20:49:09 +0000 (21:49 +0100)]
cache: change ->v_cache_dd synchronisation rules
Instead of resorting to seqc modification take advantage of immutability
of entries and check if the entry still matches after everything got
prepared.
MFC b8051298b0a3:
Fix missing value in uar_page field for ratelimit in mlx5en(4).
This is a regression issue after the new UAR API was introduced
by f8f5b459d21e .
Bjoern A. Zeeb [Mon, 25 Jan 2021 22:41:50 +0000 (22:41 +0000)]
firmware(9): extend firmware_get() by a "no warn" flag.
With the upcoming usage from LinuxKPI but also from drivers
ported natively we are seeing more probing of various
firmware (names).
Add the ability to firmware(9) to silence the
"firmware image loading/registering errors" by adding a new
firmware_get_flags() functions extending firmware_get() and
taking a flags argument as firmware_put() already does.
Requested-by: zeising (for future LinuxKPI/DRM)
Sponsored-by: The FreeBSD Foundation
Sponsored-by: Rubicon Communications, LLC ("Netgate")
MFC after: 3 days
Reviewed-by: markj
Differential Revision: https://reviews.freebsd.org/D27413
Alexander Motin [Thu, 28 Jan 2021 20:53:49 +0000 (15:53 -0500)]
Make software iSCSI more configurable.
Move software iSCSI tunables/sysctls into kern.icl.soft subtree.
Replace several hardcoded length constants there with variables.
While there, stretch the limits to better match Linux' open-iscsi
and our own initiator with new MAXPHYS of 1MB. Our CTL target is
also optimized for up to 1MB I/Os, so there is also a match now.
For Windows 10 and VMware 6.7 initiators at default settings it
should make no change, since previous limits were sufficient there.
Tests of QD1 1MB writes from FreeBSD over 10GigE link show throughput
increase by 29% on idle connection and 132% with concurrent QD8 reads.
The NS_MOREFRAG flag can be set in a netmap slot to represent a
multi-fragment packet. Only the last fragment of a packet does
not have the flag set. On TX rings, the flag may be set by the
userspace application. The kernel will look at the flag and use it
to properly set up the NIC TX descriptors.
On RX rings, the kernel may set the flag if the packet received
was split across multiple netmap buffers. The userspace application
should look at the flag to know when the packet is complete.
By default, axgbe driver does a receiver reset after predefined number
of retries for the link to come up. However, this receiver reset
doesn't always suffice, due to an hardware issue.
In that case, as a workaround, a complete phy reset is necessary.
This patch introduces a sysctl that can be set to 1 to let the driver
reset the phy completely, rather than just doing receiver reset.
The workaround will be removed once the issue is fixed by means
of firmware update.
This patch also fixes the handling of the direct attach cables
properly.
All C compilers in 2021 support standard C and architectures that did
not were retired long ago. Simplify by removing now redundant
pre-standard C code.
Jessica Clarke [Wed, 27 Jan 2021 19:19:00 +0000 (19:19 +0000)]
libllvmminimal: Add missing Support/ABIBreak.cpp
When building natively on RISC-V, linking the bootstrap clang-tblgen
fails with:
ld: error: undefined symbol: llvm::EnableABIBreakingChecks
>>> referenced by PrettyStackTrace.cpp
>>> PrettyStackTrace.o:(.sdata+0x0) in archive
/usr/obj/usr/src/freebsd-src/riscv.riscv64/tmp/obj-tools/lib/clang/libllvmminimal/libllvmminimal.a
>>> referenced by Signals.cpp
>>> Signals.o:(.sdata+0x8) in archive
/usr/obj/usr/src/freebsd-src/riscv.riscv64/tmp/obj-tools/lib/clang/libllvmminimal/libllvmminimal.a
>>> referenced by Timer.cpp
>>> Timer.o:(.sdata+0x28) in archive
/usr/obj/usr/src/freebsd-src/riscv.riscv64/tmp/obj-tools/lib/clang/libllvmminimal/libllvmminimal.a
This is likely due to Error.h's inclusion of abi-breaking.h. It's
unclear why this only affects RISC-V, but perhaps relates to its more
eager use of .sdata due to the ABI's support for linker relaxations.
Regardless, this is theoretically an issue for all architectures.
Reported by: Dennis Clarke <dclarke@blastwave.org>
A long-standing bug in Pass 1 of fsck_ffs in which it is reading in
blocks of inodes to check their block pointers. It failed to round
up the size of the read to a disk block size. When disks would
accept 512-byte aligned reads, the bug rarely manifested itself.
But many recent disks will no longer accept 512-byte aligned reads
but require 4096-byte aligned reads, so the failure to properly
round-up read sizes to multiples of 4096 bytes makes the error
much more likely to occur.
Reported by: Peter Holm and others
Tested by: Peter Holm and Rozhuk Ivan
MFC after: 3 days
Sponsored by: Netflix
Revert 2d4422e7991a, Eliminate lock order reversal in UFS ffs_unmount().
After discussion with Chuck Silvers (chs@) we have decided that
there is a better way to resolve this lock order reversal which
will be committed separately.
Kyle Evans [Wed, 27 Jan 2021 18:54:07 +0000 (12:54 -0600)]
stand: ensure that the efi directory's dependencies are correct
efi, like the various ${MACHINE} directories, should have a dependency on
the enabled interpreters.
The general rule here is that any top-level directory that has a program at
any depth within that includes loader.mk should add ${INTERP_DEPENDS} added
to its dependencies so that the appropriate ficl/lua bits are ready before
they begin.
Note that the only directories in-tree that require it but will not get it
in a more appropriate manner are i386 (on amd64), efi, and userboot. i386
and userboot are handled explicitly in Makefile.amd64 where they are added
to S.yes.
Mark Johnston [Wed, 27 Jan 2021 20:31:10 +0000 (15:31 -0500)]
qat.4: Minor tweaks
- Document a constraint on the AAD size for AES-GCM.
- Note that the list of supported platforms and add-on devices is not
complete and indicate that QAT devices will show up in pciconf
output. [1]
The reverted commit was motivated by a problem observed on stable/12,
but it turns out that a better solution was committed in r348309 but not
MFCed. So, revert this change since it is unnecessary and not really
correct: it assumes that the order in which module metadata records is
defined determines their order in the output linker set. While this
seems to hold in my testing, it is not guaranteed.
Dimitry Andric [Tue, 26 Jan 2021 13:07:47 +0000 (14:07 +0100)]
Fix clang assertion when compiling the devel/onetbb port
Merge commit 740a164de from llvm git (by Richard Smith):
PR46377: Fix dependence calculation for function types and typedef
types.
We previously did not treat a function type as dependent if it had a
parameter pack with a non-dependent type -- such a function type depends
on the arity of the pack so is dependent even though none of the
parameter types is dependent. In order to properly handle this, we now
treat pack expansion types as always being dependent types (depending on
at least the pack arity), and always canonically being pack expansion
types, even in the unusual case when the pattern is not a dependent
type. This does mean that we can have canonical types that are pack
expansions that contain no unexpanded packs, which is unfortunate but
not inaccurate.
We also previously did not treat a typedef type as
instantiation-dependent if its canonical type was not
instantiation-dependent. That's wrong because instantiation-dependence
is a property of the type sugar, not of the type; an
instantiation-dependent type can have a non-instantiation-dependent
canonical type.
Merge commit 9cf98d26e from llvm git (by Richard Smith):
PR46637: Fix handling of placeholder types in trailing-return-types.
Only permit a placeholder type in a trailing-return-type if it would
also have been permitted in the decl-specifier sequence of a
corresponding declaration with no trailing-return-type. The standard
doesn't actually say this, but this is the only thing that makes sense.
Also fix handling of an 'auto' in a trailing-return-type in a parameter
of a generic lambda. We used to crash if we saw such a thing.
Merge commit 234f51a65 from llvm git (by Richard Smith):
Don't crash if we deserialize a pack expansion type whose pattern
contains no packs.
Kyle Evans [Sun, 24 Jan 2021 19:25:34 +0000 (13:25 -0600)]
lualoader: improve loader.conf var processing
lualoader was previously not processing \ as escapes; this commit fixes
that and does better error checking on the value as well.
Additionally, loader.conf had some odd restrictions on values that make
little sense. Previously, lines like:
kernel=foo
Would simply be discarded with a malformed line complaint you might not
see unless you disable beastie.
lualoader tries to process these as well as it can and manipulates the
environment, while forthloader did minimal processing and constructed a
`set` command to do the heavy lifting instead. The lua approach was
re-envisioned from building a `set` command so that we can appropriately
reset the environment when, for example, boot environments change.
Lift the previous restrictions to allow unquoted values on the right hand
side of an expression. Note that an unquoted value is effectively:
[A-Za-z0-9-][A-Za-z0-9-_.]*
This commit also stops trying to weirdly limit what it can handle in a
quoted value. Previously it only allowed spaces, alphanumeric, and
punctuation, which is kind of weird. Change it here to grab as much as it
can between two sets of quotes, then let processEnvVar() do the needful and
complain if it finds something malformed looking.
My extremely sophisticated test suite is as follows:
Future work may provide a stub loader module in userland so that we can
formally test the loader scripts rather than sketchy setups like the above
in conjunction with the lua-* tools in ^/tools/boot.
Kyle Evans [Wed, 20 Jan 2021 14:01:25 +0000 (08:01 -0600)]
build: remove LIBPTHREAD/LIBTHR build options
WITHOUT_LIBTHR has been broken for a little over five years now, since the
xz 5.2.0 update introduced a hard liblzma dependency on libthr, and building
a useful system without threading support is becoming increasingly more
difficult.
Additionally, in the five plus years that it's been broken more reverse
dependencies have cropped up in libzstd, libsqlite3, and libcrypto (among
others) that make it more and more difficult to reconcile the effort needed
to fix these options.
Ryan Moeller [Wed, 27 Jan 2021 19:27:46 +0000 (14:27 -0500)]
sbin/sysctl: Fix CTLFLAG_SKIP for adjacent nodes
The OID is saved when we encounter CTLFLAG_SKIP so that descendants can
be skipped as well. We then must not update the skip OID until we are
out of the node. This was achieved by resetting the skip OID once the
prefix no longer matches, but the case where the OID we reset on has
CTLFLAG_SKIP was not accounted for.
Ed Maste [Fri, 22 Jan 2021 17:22:35 +0000 (12:22 -0500)]
elfctl: allow features to be specified by value
This will allow elfctl on older releases to set bits that are not yet
known there, so that the binary will have the correct settings applied
if run on a later FreeBSD version.
PR: 252629 (related)
Suggested by: kib
Reviewed by: gbe (manpage, earlier), kib
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D28284
Target value for val has uint32_t type, not uint, adjust used constant.
Change val type to unsigned so that left and right sides of comparision
operator do not expose different signed types of same range [*].
Switch to unsigned long long and strtoll(3) so that 0x80000000 is
accepted by conversion function [**].
Reported by: kargl [*]
Noted by: emaste [**]
Reviewed by: emaste
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D28301
Kyle Evans [Fri, 22 Jan 2021 23:13:42 +0000 (17:13 -0600)]
jemalloc: restore JEMALLOC_DEBUG when building WITHOUT_MALLOC_PRODUCTION
The default for MALLOC_PRODUCTION was switched to ON in 02611ef8ee9f. This
effectively reverts the jemalloc header change from bfd15705156b so that
the knob behaves exactly as it does on a -CURRENT; that is, we are
effectively doing:
WITH_MALLOC_PRODUCTION -> -DMALLOC_PRODUCTION (default for stable/* and on)
WITHOUT_MALLOC_PRODUCTION -> -DJEMALLOC_DEBUG (default for main)
This allows the knob to be used for debugging on stable branches, too, which
is believed to be the main reason one would want to twist it off.