LinuxKPI is asking for single-segment mappings. Some (wireless) drivers
are using this to map multi-pages and our busdma framework is not very
friendly to that as single-segments [D31823]. Add a counter so we can
track when this happens to gather more information.
Sponsored by: The FreeBSD Foundation
Reviewed by: hselasky
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D34715
This is a temporary hack for zlib to make sure that the library
still builds when building with Z_SOLO (used in kernel and loader),
as zlib is depending on limits.h which is only available in STDC
case.
PR: kern/262977
MFC after: 3 days
As of today, using 'kldload miibus' is not equivalent to using 'device
miibus' in a kernel config. Newly introduced PHY drivers (DP83822,
DP83867, VSCPHY) and source files/PHY driver for FDT-enabled kernels
are missing. Without including them, kernel modules using any function
from dev/mii/mii_fdt.c refuse to load. Additionally, miivar.h directly
includes opt_platform.h.
Add the missing sources to the module build, with the FDT-only files
gated behind an OPT_FDT check. Maintain the alphabetical listing of
SRCS, but move the required header files to a separate line to improve
readability.
Reviewed by: mhorne, mindal@semihalf.com
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D34256
For development, building a driver as kernel module is both convenient
and a time saver (no need for reboot on some change, testing it requires
just kldunload and kldload, a matter of seconds). For some special
cases, it may be even desirable to postpone initializing the network
interface after some action is done (loading a FPGA bitstream may be
required for Zynq/ZynqMP based hardware as an example).
Building is limited to ARM, ARM64 and RISC-V architectures (for Zynq,
ZynqMP, PolarFire Soc based boards, and HiFive based boards are known to
use CGEM at the moment).
Reviewed by: mhorne
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D34687
In some cases vn_open_cred overwrites cn_flags, effectively nullifying
initialisation done in NDINIT. This will have to be fixed.
In the meantime make sure the flag is passed.
Reported by: jenkins
Noted by: Mathieu <sigsys@gmail.com>
When trying to avoid a CARP demotion during a pfsync service restart, I
noticed that a non-default value for the net.pfsync.carp_demotion_factor
sysctl was not being applied during the demotion. The CARP was always
demoted by 240.
After investigating, I realized that the sysctl was using VNET_NAME()
without the CTLFLAG_VNET.
PR: 262983
Reviewed by: kp
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D34737
As in 4a22cd6c4e nf and rss should be
signed and not unsigned. Change the types in the header and while
here change a magic number to a define as done elsewhere (value does
not change).
When calculating c_rssi we need to make it relative so subtract nf.
And while here improve the debug output.
This will hopefully fix ifconfig wlanN list scan S:N output which
tools use to chose a BSSID and help net80211 internal calculations.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
This fixes userspace RDMA applications that would fail due to mmap
failure. The driver's mmap routine would succeed but later the
linux_compat.c mmap routine would fail because vma->vm_private_data
wasn't set properly.
This is catch-up with b633e08c70.
Reported by: Veeresh @ Chelsio
Sponsored by: Chelsio Communications
In the changes to get hystart++ into cubic an inadvertent line
was removed in the conditional to figure out if you need to exit
hystart++ back to slowstart. The line of course is the most crucial
one (the others are valid but not critical) i.e. is the new rtt
less than the point where we entered hystart++. Without the line
we end up bouncing in and out of CSS.
Reported By: Reese Enghardt
Sponsored By: Netflix Inc.
There is a case where rack will get stuck when it has outstanding data and
the peer collapses the rwnd down to 0. This leaves the session hung if
the rwnd update is not received. You can test this with the packet drill script
below. Without this fix it will be stuck and hang. With it we retransmit everything.
This also fixes the mtu retransmit case so we don't go into recovery when
the mtu is changed to a smaller value.
Reviewed by: tuexen
Sponsored by: Netflix Inc.
Differential Revision: https://reviews.freebsd.org/D34573
Add cap_lo and cap_hi sysctl to each nvme drive. This publishes the raw
capabilities of the drive. Now we can only discover these with
bootverbose.
Sponsored by: Netflix
Setting MPS in the CC should be a power of 2 number (it specifies the
page size of the host is 2^(12+MPS)), so adjust the calcuation. There is
no functional change because we do not support any architecutres != 4k
pages (yet). Other changes are needed for architectures with 16k or 64k
pages, especially when the underlying NVMe drive doesn't support that
page size (Most drives support a range that's small, and many only
support 4k), but let's at least do this calculation correctly. 12 - 12
is just as much 0 as 4096 >> 13 is :)
Sponsored by: Netflix
Reviewed by: mav
Differential Revision: https://reviews.freebsd.org/D34707
Add man pages for rtw88 and rtw88fw. Install a copy of the firmware
license file and hook up the driver and firmware modules to the build.
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Relnotes: yes
Import the most recent versions of the firmware images for the
rtw88 driver.
This is based on linux-firmware at 681281e49fb6778831370e5d94e6e1d97f0752d6.
The license of the firmware matches the previous rtwnfw(4) firmware
files (modulo a Copyright year) and you can find a copy in
sys/contrib/dev/rtw88fw/LICENCE.rtlwifi_firmware.txt.
Add build infrastructure to create the .ko files but do not yet hook
it up to the build until all parts are in the tree.
Approved by: core (imp)
MFC after: 2 weeks
Import rtw88 based on wireless-testing at
5d5d68bcff1f7ff27ba0b938a4df5849849b47e3 with adjustments for FreeBSD.
While our version of the driver has knowledge about the incapablity
of DMA above 4GB we do see errors if people have more than that
often already showing when laoding firmware.
The problem for that is currently believed to be outside this driver
so importing it anyway for now.
Given the lack of full license texts on non-local files this is
imported under the draft policy for handling SPDX files (D29226). [1]
Approved by: core (imp) [1]
MFC after: 2 weeks
FreeBSD detects serial ports twice: First, very early in the boot
process, in order to obtain a usable console; and second, during
the device probe/attach process. When a UART is discovered during
device probing, FreeBSD attempts to determine whether it is a
device which was already being used as a console; without this,
the console doesn't work in userland.
Unfortunately it's possible for a UART to be mapped to a different
location in memory when it is discovered on a bus than it has when
it is announced via the ACPI SPCR table; this breaks the matching
process, which relies on comparing bus addresses.
To address this, we introduce a concept of "unique" serial devices,
i.e. devices which are guaranteed to be present *only once* on any
system. If we discover one of these during device probing, we can
match it to a same-PCI-vendor-and-device-numbers console which was
announced via the ACPI SPCR table, regardless of the differing bus
addresses.
At present, the only unique serial device is the "Amazon PCI serial
device" (vendor 0x1d0f, device 0x8250) found in some EC2 instances.
This unbreaks the serial console on those systems.
Reviewed by: imp
MFC after: 3 days
Sponsored by: https://www.patreon.com/cperciva
Differential Revision: https://reviews.freebsd.org/D34703
In vm_phys_alloc_queues_contig, in the case that a sequence of
max-order blocks are sought to fulfill an allocation, a sequence is
ruled out if it does not have enough max-order blocks to satisfy the
allocation. However, there may be smaller blocks of free memory that
follow the last max-order block in the sequence, and they may be big
enough to complete the allocation request, so check for that
possibility before giving up on that block sequence.
Reviewed by: markj
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D34724
stge_attach() could fail at line 464, sc->sc_spec remains NULL when
calling stge_detach(), thus bus_release_resources() at line 704 will
trigger null pointer dereference. We need to check the nulliness before
calling bus_release_resources().
PR: 258420
Reviewed by: markj
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D34629
To avoid annoyng messages from LTP test suites add the simple
implementation of /proc/self/oom_score_adj which is do nothing.
Reviewed by: emaste
Differential revision: https://reviews.freebsd.org/D34710
MFC after: 2 weeks
Historically 32-bit Linuxulator under amd64 emulated the real i386
behavior. Since 3d8dd983 the old i386 Linux world can't be used under
amd64 Linuxulator as it don't know anything about amd64 machine (which
is returned now by newuname() syscall). So, add a knob to allow to swith
the behavior and use i386 Linux binaries on amd64.
Set knob to the new behavior as I think this is common to the modern
Linux distros.
Reviewed by: Pau Amma (doc), emaste
Differential revision: https://reviews.freebsd.org/D34708
MFC after: 2 weeks
Avoid calculating d_off value as it is specific to the underlying filesystem
and can be used by others API, like lseek(), seekdir() as input offset.
Differential revision: https://reviews.freebsd.org/D31551
MFC after: 2 weeks
Since Linux 5.4, if id is zero, then wait for any child that is in the same
process grop as the caller's process group.
Differential revision: https://reviews.freebsd.org/D31567
MFC after: 2 weeks
As FreeBSD does not have __WALL option bit analogue explicitly set all
possible option bits to emulate Linux __WALL wait option bit.
Reviewed by: emaste
Differential revision: https://reviews.freebsd.org/D31555
MFC after: 2 weeks
Compiling another driver on i386 revealed two problems:
- ieee80211_tx_info.status.status_driver_data space needs to be
calculated. While a pointer is 32bit vm_paddr_t is 64 bit on i386
so we didn't fit more than one of these in but needed more space.
- the arguments to ieee80211_txq_get_depth() are expected to
unsigned long and not uint64_t.
No user noticable changes.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
In the search for contiguous pages, as each page segment is examined,
check to see if the free list set for the next page segment differs
from the set for the current segment, and avoid a pointless search if
they do not differ.
Discussed with: alc
Reviewed by: markj
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D33947
Diff reduction between mpr and mps.
Fixes: e2997a03b7 ("Diagnostic buffer fixes for the mps(4)...")
MFC after: 3 days
Sponsored by: The FreeBSD Foundation
The limit may later be updated by the "set limit" directive in pf.conf.
UMA does not permit a limit to be set on a zone after any items have
been allocated from a zone.
Other UMA zones used by pf do not appear to be susceptible to this
problem: they either set a limit at zone creation time or never set one
at all.
PR: 260406
Reviewed by: kp
MFC after: 3 days
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34713
In vmbus_pcib_prepopulate_bars(), after writing all 1's to the
avialable device bars, those without being configured by device driver
are also set to its initialized values. However, this could cause
weird problem which results to device failure. The issue has been
reported to happen on LSI 9211-8i HBA card for DDA access on Hyper-V.
Writing back the orignal BAR values seem to work around this problem.
Reported by: Alexander Motin <mavbsd@gmail.com>
Tested by: Mathias Kraut <krautmaster@gmail.com>
Fixes: 75412a521f Hyper-V: vPCI: Prepopulate device bars
MFC after: 1 month
32-bit architectures other than i386 have 64-bit time_t which results
in a struct timespec with 12 bytes for tv_sec and tv_nsec, and 4 bytes
of padding. Zero the padding holes in struct stat32 and struct
freebsd11_stat32.
i386 has 32-bit time_t; struct timespec is 8 bytes and has no padding.
Found by inspection, prompted by a report by Reno Robert of Trend Micro
Zero Day Initiative. The originally reported issue (ZDI-CAN-14538) is
already fixed in all supported FreeBSD versions (it was addressed
incidentally as part of the 64-bit inode project).
Reviewed by: markj
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34709
Possibility to do it was always a bug, but it runs into crashes
since recent introduction of a per-ruleset RB tree.
Reviewed by: kp
Sponsored by: Rubicon Communications, LLC ("Netgate")
Reported by: syzbot+665b700afc6f69f1766a@syzkaller.appspotmail.com
Variants of for_each_sg/for_each_sg_dma_page but they operate on sgtable
structs.
Needed by drm v5.10
MFC after: 1 week
Sponsored by: Beckhoff Automation GmbH & Co. KG
*_CFG_PAGE ioctl handlers in the mpr, mps, and mpt drivers allocated a
buffer of a caller-specified size, but copied to it a fixed size header.
Add checks that the size is at least the required minimum.
Note that the device nodes are owned by root:operator with 0640
permissions so the ioctls are not available to unprivileged users.
This change includes suggestions from scottl, markj and mav.
Two of the mpt cases were reported by Lucas Leong (@_wmliang_) of
Trend Micro Zero Day Initiative; scottl reported the third case in mpt.
Same issue found in mpr and mps after discussion with imp.
Reported by: Lucas Leong (@_wmliang_), Trend Micro Zero Day Initiative
Reviewed by: imp, mav
MFC after: 3 days
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34692
Commit 867c27c23a modified the NFS client so that
it did IO_APPEND writes directly to the NFS server
bypassing the buffer cache, via a call to
nfs_directio_write(). Unfortunately, this (very old)
function assumed that the uio iov was for user space
addresses. As such, a IO_APPEND VOP_WRITE() that
was for system space, such as ktrace(1) does, would
write bogus data.
This patch fixes nfs_directio_write() so that it
handles kernel space uio iovs.
Reported by: bz
Tested by: bz
MFC after: 2 weeks
This more clearly differentiates system call arguments from integer
registers and return values. On current architectures it has no effect,
but on architectures where pointers are not integers (CHERI) and may
not even share registers (CHERI-MIPS) it is necessiary to differentiate
between system call arguments (syscallarg_t) and integer register values
(register_t).
Obtained from: CheriBSD
Reviewed by: imp, kib
Differential Revision: https://reviews.freebsd.org/D33780
Some filesystems do not fill out certain optional vattr fields. To
ensure that they do not get copied out to userspace uninitialized, use
VATTR_NULL to provide default values.
Reported by: KMSAN
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Some new AMD systems provide a HPET MMIO region smaller than the 1KB
specified, and a correspondingly small number of timers. Handle this in
the HPET driver rather than requiring a 1KB window. This allows the
HPET driver to attach on such systems.
PR: 262638
Reviewed by: markj
MFC after: 1 month
with md5 sum used as key.
This gets rid of the quadratic rule traversal when "keep_counters" is
set.
Reviewed by: kp
Sponsored by: Rubicon Communications, LLC ("Netgate")
Makes it cheaper to compare rules when "keep_counters" is set.
This also sets up keeping them in a RB tree.
Reviewed by: kp
Sponsored by: Rubicon Communications, LLC ("Netgate")
For now only protects rule creation/destruction, but will allow
gradually reducing the scope of rules lock when changing the
rules.
Reviewed by: kp
Sponsored by: Rubicon Communications, LLC ("Netgate")
Otherwise all anchors hash to the same value.
Note this can result in checksum mismatches between pfsynced hosts,
but it has to be sorted out as the previously computed checksum
would fail to indicate changed anchors.
Reviewed by: kp
Sponsored by: Rubicon Communications, LLC ("Netgate")
Make sure both sides of a comparison are unsigned. As the values being
compared are size_t make the the value in the for loop size_t too.
Sponsored by: The FreeBSD Foundation
On arm64 we can ask the hardware to perform cache operations from
userspace. These require read permission however when the memory is
unmapped the kernel will receive a write exception. Add a check to
see if the cause of the exception is from the cache and pass a memory
read fault type to the vm subsystem.
PR: 262836
Reported by: dch
Sponsored by: The FreeBSD Foundation
Use of stdatomic.h is undefined in C++, even the C++ 2020 standard does not
list stdatomic.h as a C library header supported by the language. More,
there are some subtle differences between the <atomic> C++ header, and
C11+ stdatomic.h provided features.
Nonetheless, it is a quality of the implementation aspect, so let mis-users
mis-use stdatomic.h as they want, by making a compat shim for _Bool.
PR: 262683
Reported by: yuri
Reviewed by: dim, emaste
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D34686
This completes the patch which was originally meant to go in.
Spotted by: mhorne
Fixes: c35ec1efdc ("vfs: [1/2] fix stalls in vnode reclaim by not
requeieing from vnlru")
It was only called in the non-NUMA and single-domain paths.
Some of its assertions were duplicated in uma_zalloc_domain,
but some things were missed, especially memguard.
Reviewed by: markj, rstone
MFC after: 1 week
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D34472
The uma_zalloc functions expect exactly one of [M_NOWAIT, M_WAITOK].
If neither or both are passed, print an error and a stack dump.
Only do this ten times, to prevent livelock. In the future, after
this exposes enough bad callers, this will be changed to a KASSERT().
Reviewed by: rstone, markj
MFC after: 1 month
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D34452
It expects exactly one of those flags. A future commit will assert this.
Reviewed by: rstone
MFC after: 1 month
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D34451
uma_zalloc_arg expects exactly one of the two WAIT flags. A future
commit will assert this.
Reviewed by: rstone
MFC after: 1 month
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D34450
Turns out execve looks at it to store binary name, but in order to
trigger the problem one has to be trying to exec '/'. As is the value
would be left uninitialized (or rather set to -1 on debug kernels).
Fixes: 56244d3574 ("vfs: hoist degenerate path lookups out of the
loop")
This can leak kernel stack data otherwise.
Reviewed by: imp, markj
Sponsored by: The University of Cambridge, Google, Inc.
Differential Revision: https://reviews.freebsd.org/D34594
Following ARMARM sec D5.2.11, which says:
> Where an instruction results in an update to a System register,
> as is the case with the AT * address translation instructions,
> explicit synchronization must be performed before the result is
> guaranteed to be visible to subsequent direct reads of the
> PAR_EL1.
Reviewed By: andrew
MFC after: 3 weeks
Sponsored by: Ampere Computing
Differential Revision: https://reviews.freebsd.org/D34665
Bits 43:0 of the TLBI operand are bits 55:12 of the VA. Leaving
bits 63:55 of the VA in bits 51:44 of the operand might wind up
setting the TTL field (47:44) and accidentally restricting which
translation levels are flushed in the TLB.
Reviewed By: andrew
MFC after: 3 days
Sponsored by: Ampere Computing
Differential Revision: https://reviews.freebsd.org/D34664
Specific to Linux AT_NO_AUTOMOUNT flag tells the kernel to not automount the
terminal component of pathname if it is a directory that is an automount point.
As it is the default for FreeBSD silencly ignore this flag.
glibc-2.34 uses this flag in the stat64 system calls which is used by i386.
Reviewed by: trasz
Differential revision: https://reviews.freebsd.org/D31524
MFC after: 2 weeks
Disallow the use of tables in ethernet rules. Using tables requires
taking the PF_RULES lock. Moreover, the current table code isn't ready
to deal with ethernet rules.
Disallow their use for now.
Sponsored by: Rubicon Communications, LLC ("Netgate")
The firmware doesn't report FORCE_FEC in pcaps if the transceiver
plugged in at that time does not support a speed that may use FEC. It
is incorrect for the driver to assume that the FORCE_FEC value it read
during attach (in init_link_config) is permanent. Instead, it should
check pcaps just before issuing the L1CFG command.
MFC after: 1 week
Sponsored by: Chelsio Communications
Cleanup some debugging. Rename the global variable to be less
generic. Hide all debugging behind #ifdef for now and turn off.
Rename the debugging sysctl so we can start adding more to the
subtree.
There is a need to change that wildly grown infrastructure into
something more homogenic soon but this should do for 13.1.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
This register set contains the values of the fsbase and gsbase
registers. Note that these registers can already be controlled
individually via ptrace(2) via MD operations, so the main reason for
adding this is to include these register values in core dumps. In
particular, this will enable looking up the value of TLS variables
from core dumps in gdb.
The value of NT_X86_SEGBASES was chosen to match the value of
NT_386_TLS on Linux. The notes serve similar purposes, but FreeBSD
will never dump a note equivalent to NT_386_TLS (which dumps a single
segment descriptor rather than a pair of addresses) and picking a
currently-unused value in the NT_X86_* range could result in a future
conflict.
Reviewed by: markj
Differential Revision: https://reviews.freebsd.org/D34650
Extending what was started in d3ef7fb459,
when a driver signals that hw_scan is not possible and expects a sw_scan
to be preformed we triggered a sw_scan towards the driver but did not
let net80211 know.
Cancel the initial scan towards net80211. If we we defer to sw_scan
then clear IEEE80211_FEXT_SCAN_OFFLOAD so net80211 will send probe
requests, and actively start a new scan with net80211.
This may have to be further refined in the future but seems to work
for the moment.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Remove the originally disabling of (*ic_scan_curchan), which for iwlwifi
was not needed. The condition always only was approximate.
A set IEEE80211_FEXT_SCAN_OFFLOAD will still prevent net80211 from sending
probe_reqs if handled by driver/firmware.
ic_scan_curchan will re-arm the timer to switch channels for drivers which
need it (e.g., rtw88, but that again is a NOP for iwlwifi).
So enabling ic_scan_curchan should not have further side effects for iwlwifi
but allow other drivers to work better.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
When checking for IEEE80211_FEXT_SCAN_OFFLOAD, do so on the vap rather
than ic. This brings us in line with what net80211 does.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
For (*config)() based drivers not using chanctx we need to use
ic_set_channel() to switch channels. So far this was disabled
based on scanning flags (as swscan is one of the initial use cases
for this function). Now make it only dependent on (*config)()
for the moment to save us the work if (*config)() is not supported.
For iwlwifi (*config)() is a NOP so no functional changes there
but for other drivers such as rtw88 this will allow us to scan and
set the channel (which helps to receive on channels other than 1).
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
The keyboard driver was initially protected via spl* interrupt priority
calls but (as part of a comprehensive effort) migrated to use the Giant
lock (mutex).
The spl calls left behind became NOPs but they can be confusing as they
have no bearing on the actual mutual exclusion that is now present.
Remove them from kbd and add assertions that Giant is held. markj notes
that there is conflation between the "bus topo" lock (which is Giant
under the hood) and Giant. The assertions could either be addressed as
a small item along with bus topology locking work or they'll be removed
if kbd is decoupled from Giant.
PR: 206680
Reviewed by: markj
MFC after: 3 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34645
Rather than manually setting up a chandef and then effectively forcing
a memcpy, use cfg80211_chandef_create() to do the work for us entirely.
This works here as we do not store the resulting chandef separately
for other use.
While here remove a duplicate assignment in cfg80211_chandef_create().
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
If a driver does not support (*sta_state)() we internally in
lkpi_80211_mo_sta_state() fall back to using (*sta_add/*sta_remove)().
In that case add tracking of both added_to_drv and state fields for the
lsta so that our state machine keeps working and assertions do not fire.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
This permits I/O devices on the host to directly access wired memory
dedicated to guests using passthru devices. Note that wired memory
belonging to guests that do not use passthru devices has always been
accessible by I/O devices on the host.
bhyve maps guest physical addresses into the user address space of
the bhyve process by mmap'ing /dev/vmm/<vmname>. Device models pass
pointers derived from this mapping directly to system calls such as
preadv() to minimize copies when emulating DMA. If the backing store
for a device model is a raw host device (e.g. when exporting a raw disk
device such as /dev/ada<n> as a drive in the guest), the host device
driver (e.g. ahci for /dev/ada<n>) can itself use DMA on the host
directly to the guest's memory. However, if the guest's memory is
not present in the host IOMMU domain, these DMA requests by the host
device will fail without raising an error visible to the host device
driver or to the guest resulting in non-working I/O in the guest.
It is unclear why guest addresses were removed from the IOMMU host domain
initially, especially only for VM's with a passthru device as the
host IOMMU domain does not affect the permissions of passthru devices,
only devices on the host.
A considered alternative was using bounce buffers instead (D34535
is a proof of concept), but that adds additional overhead for unclear
benefit.
This solves a long-standing problem when using passthru devices and
physical disks in the same VM.
Thanks to: grehan (patience and help)
Thanks to: jhb (for improving the commit message)
PR: 260178
Reviewed by: grehan, jhb
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D34607
Allow a driver to overwrite the bsddriver name (we use for pci and
for wlan parent devices). This allows us to to set
.bsddriver.name in struct pci_driver passed to module_pci_driver()
and with that set the BSD driver name while retaining the Linux .name
one.
This is helpful for divers which have different parts depending on
chipset and with that would change driver names which is highly
confusing especially for configuration. One example is an upcoming
rtw88 driver which would be rtw_8822be or rtw_8822ce depending on
chipset.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Reviewed by: emaste, hselasky
Differential Revision: https://reviews.freebsd.org/D34653
PCM_REGISTERED(d) tests that d is not NULL, so perform that check first
as we may have cases where devclass_get_softc has a null entry.
PR: 262671
Reviewed by: hselasky
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34648
This register set exposes the per-thread TLS register. It matches the
layout used by Linux on arm64. Linux does not implement this note for
32-bit arm.
Reviewed by: andrew, markj
Sponsored by: University of Cambridge, Google, Inc.
Differential Revision: https://reviews.freebsd.org/D34595
This includes adding support for NT_ARM_VFP for 32-bit binaries
running under aarch64 kernels both for ptrace(), and coredumps via the
kernel and gcore.
Reviewed by: andrew, markj
Sponsored by: University of Cambridge, Google, Inc.
Differential Revision: https://reviews.freebsd.org/D34448
Similar to fill_fpregs(), only invoke vfp_save_state() for curthread.
While here, zero the buffer if FP hasn't been started to avoid leaking
kernel stack memory.
Reviewed by: andrew, markj
Sponsored by: University of Cambridge, Google, Inc.
Differential Revision: https://reviews.freebsd.org/D34525
This permits adding a register set for FREEBSD32 ABIs.
While here, include <sys/linker_set.h> to make this header more
self-contained in the kernel.
Reviewed by: andrew, markj
Sponsored by: The University of Cambridge, Google, Inc.
Differential Revision: https://reviews.freebsd.org/D34536
This has been safe since e9e7bc8250, which moved parts of error
handling from the ithread to a taskqueue.
MFC after: 1 week
Sponsored by: Chelsio Communications
Fix two last-minute changes of d9f59799fc:
(1) for consitency name the debugging function lkpi_lsta_dump()
(2) pass in the original node (ni) rather than taking it from the lsta
to avoid problems realted to bss_update as mentioned in the above
commit.
No functional changes for users.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
X-MFC with: d9f59799fc
In skb_queue_splice_init() we set a next value and then used that new
value to further update the remaining linking rather than the original
value. Introduce another temporary variable 'n' to hold the original
value and use that.
While here rename q and h to from and to as otherwise it was too
confusing to read.
Also initialize skb->prev and skb->next to point to skb itself if
for nothing else at least to aid debugging.
Reported by: phk (panic in iwl_txq_reclaim)
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
We use the p_itcallout callout, interlocked by the proc lock, to
schedule timeouts for the setitimer(2) system call. When a process
exits, the callout must be stopped before the process struct is
recycled.
Currently we attempt to stop the callout in exit1() with the call
_callout_stop_safe(&p->p_itcallout, CS_EXECUTING). If this call returns
0, then we sleep in order to drain the callout. However, this happens
only if the callout is not scheduled at all. If the callout thread is
blocked on the proc lock, then exit1() will not block and the callout
may execute after the process has fully exited, typically resulting in a
panic.
I cannot see a reason to use the CS_EXECUTING flag here. Instead, use
the regular callout_stop()/callout_drain() dance to halt the callout.
Reported by: ler
Tested by: ler, pho
MFC after: 1 month
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34625
It's unneeded as it was just used to align KERNBASE to a level 2
block start address. KERNBASE was already aligned correctly.
Sponsored by: The FreeBSD Foundation
When the page size the kernel is built for is not the same as
EFI_PAGE_SIZE we need to increment the page index at a faster rate.
Add this adjustment to the arm64 EFIRT support in preperation for
experimental 16k PAGE_SIZE support.
Sponsored by: The FreeBSD Foundation
To support cc -pg on arm64 we need to implement .mcount. As clang and
gcc think it is function like it just needs to load the arguments
to _mcount and call it.
On gcc the first argument is passed in x0, however this is missing on
clang so we need to load it from the stack. As it's the caller return
address this will be at a known location.
PR: 262709
Reviewed by: emaste (earlier version)
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34634
If we need to resched it takes the lock, resched, release the lock
and returns 1, otherwise simply returns 0.
Needed by drm v5.9
MFC after: 1 month
Reviewed by: hselasky
Sponsored by: Beckhoff Automation GmbH & Co. KG
Differential Revision: https://reviews.freebsd.org/D34620
Rework the state machine parts for various reasons:
(1) to add sta tracing to be able to better follow ni and lsta state
(2) factor out/implement lkpi_lsta_remove() to unlink the lsta and
free the ni reference.
(3) avoid calling lkpi_disassoc() when you would think you should as
changing BSS_CHANGED_ASSOC setting vif->bss_conf.assoc to false
triggers a sta removal from firmware in iwlwifi which then triggers
follow-up errors. I do not understand why they use flags and state
and ?? in parallel (too many options and ways to do things?).
(4) when "roaming" (or being disassoc/deauth) from an AP both net80211
and apparently so mac80211 re-start with a new node/sta. This
results in us losing one or the other state in the compat layer
or not updating firmware appropriately. To resolve this make use
of (a) the newly introduced (*iv_update_bss)() and (b) always tear
a station down to "State 1" (INIT/SCAN/pre-AUTH) and only if needed
re-create the new one (if we go to AUTH).
A slightly earlier version has survived a night of wpa_supplicant
and hostapd fighting each other over disassoc and deauth and
re-associating/authorizing.
While there update a few comments and typos and do a few minor auxiliary
changes which are hard or not worth to extract.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Factor out dtim/tsf setting/updates into lkpi_update_dtim_tsf()
with tracing and add an extra update call.
This avoids some code duplication and puts maintainance into a
single place.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Split lkpi_ic_node_alloc() into two functions to make the code simpler
and to allow lkpi_lsta_alloc() to be re-used from another part of the
code related to (*iv_update_bss)().
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Implement ieee80211_beacon_loss() similar to
ieee80211_connection_loss() with different state handling.
While here leave a comment in connection_loss() about the state
change argument.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Introduce (*iv_update_bss)() with a default implementation to allow
drivers to overload/intercept the time when we swap iv_bss.
This helps firmware based drivers to synchronize state with firmware.
Otherwise, for some state changes, we begin with one ni (and in
LinuxKPI lsta) and try to finish with another ni (and a new lsta
in different state) and may no longer have access to the previous state.
This also saves us from constantly checking for ni changes complicating
code.
No functional changes intended.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
X-MFC: move (*iv_update_bss) to spare area
When forcing DEUATH in ieee80211_sta_join1() log the current state
we are coming from as well. Note this isn't always the state we
are expecting as iv_state was updated already, so contrary to the
comment we usually do not see RUN there.
Leave a comment earlier with regards to this as well.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
The vif structure includes fields at the end which are #ifdef KERNEL,
causing a mismatch between the structure sizes between kernel and
user level. netstat -g failed with an ENOMEM on the sysctl to fetch
the vif table. Change the vif sysctl code in ip_mroute to copy out
only the user-level-visible portion of each table entry.
Reviewed by: bz, wma
Differential Revision: https://reviews.freebsd.org/D34627