In _thr_once, if init_routine fails with exception, the stack allocate
__cup will become invalid when unwinder walk back and free the stack of
thr_once. This cause cleanup linked list be dangling. In here, we give
the cleanup function a landing pad so that we td can correctly point
to previous cleanup function.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D61212
_thr_check_init() is called on hot paths such as pthread_once(), and
the uninitialized case can only happen once. Mark it with
__predict_false so the compiler lays out the initialized case as the
fall-through path.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D61213
For a pshare key, we can actually cache the last one as a fast path
instead of trapping into kernel directly. It makes sense as a lock and
unlock should be a pair on the same key and at the same time. For
example, lock(A) -> unlock(A) -> lock(B) -> unlock(B). Each unlock is
a hit and refresh later (50% hit).
For a single pshare lock, it is evne a 100% hit as the key never
changed.
For nested lock lock(A) -> lock(B) -> lock(C) -> unlock(C) ->
unlock(B) -> unlock(A). It only cache the last one. However, we should
really think about rewritting this logic as it is a pshare lock shared
by many processes and nested should be uncommon.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60957
_citrus_iconv_std_iconv_convert can return EILSEQ without setting
invlen. As a result, we should check EILSEQ in the return value or it
will incorrectly return -2
Reviewed by: kevans
Fixes: e2030ca246 ("libc: fix c*rtomb/mbrtoc*")
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D61151
Only the default PSI IDs are supported, other controllers fall back to
USB 3.0. Record the speed IDs the controller has: the root hub BOS
lists those, and a link without one uses the closest lower ID.
Per xHCI 6.2.2, take the parent hub slot ID and port number of SS and
SSP devices from the closest higher-rank hub.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D52347
An SSP hub reports its sublink speeds in the BOS descriptor, and
Extended Port Status tells which one each port uses. Read it based on
bcdUSB, since the ports of a hub can be faster than its upstream link.
Without it, the ports are treated as SuperSpeed.
Read the speed again in usbd_req_re_enumerate(), since a link can come
back at another speed.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57204
Add the SuperSpeedPlus USB Device Capability descriptor and its field
macros.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57203
Get the header or the complete BOS descriptor of SS+ devices.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57202
usbd_req_get_ext_port_status() wraps the Get Extended Port Status
request.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57201
Extended Port Status reports the speed and lane count of SS and SSP
links.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57200
Handle the new speeds wherever USB_SPEED_SUPER was checked. Most
consumers treat them as SS.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57198
libusb has one speed for all SSP links, so print "SUPER+ (>=10Gbps)".
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57196
Report the new kernel speeds as LIBUSB20_SPEED_SUPER_PLUS.
Reviewed by: bz
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57195
Add the USB 3.1 and 3.2 speeds and revisions. The speeds are in USB
3.2 rank order, so they can be compared directly.
Also, move USB_*_MAX into the enums.
Reviewed by: bz
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57194
We cannot actually rely on a status (finish) before thr_exit. The
thread needs the kernel signal status=1 to userspace. Before that,
any tryjoin fails. The previous one works as we sleep 1 and should be
enough for normal case.
Reviewed by: kib, markj
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60832
thr_try_gc unlock and lock again to make sure the lock order. However,
there is no reason for such roundtrip if the caller holds the correct
oder. As a result, we create a new helper function to add the thread
to gc list without unlock + lock and wire the join_common to this new helper.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60717
The while loop previous can only be break when ret == ETIMEDOUT. They
never produce EBUSY. Remove the deadcode here.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60711
linux stores mount id as a int32_t, using xor would make it becomes
negative in some cases. As a result, we cap the value to INT_MAX to
prevent getting a negative value from userspace.
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D59202
LIBPTHREAD_QUEUE_FIFO is used in bitwise offset, we set the limit to
0~31 so that it won't cause undefined behavior. Also, create a new
helper function to do getenv++strtol directly and check the boundary of
the valid value.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60487
Like yieldloops, we shoulde be able to set _thr_spinloops to zero.
Originally, it makes us to enformce default spin time even if we try to
disable it. Make MUTEX_ADAPTIVE_SPINS a one time initialization now.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60486
A single page invalidation sets addr2 == 0. In the original code, it
falsely flush the whole address in non global pmap. However, the
kernel pmap are all PG_G, which means a single page flush will always be
staled and thus become invalid. As a result, we set parameter based on
their op in a new helper function instead of relying on args. This
affects only on AMD platform as Intel has their PTI implementation.
PR: 291577
Tested by: [email protected]
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60380
1. When contester > 2, a release causes kernel set CONTENTION bit but
with no owner. In this case, the userspace should be able to try the
lock instead of fall immediately. We copy this from trylock2.
2. A PI mutex should consult to kernel instead of decided by userspace
itself.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60427
On a 64 byte cache line CPU (amd64), m_owner is contended by all of
the shared processes. However, after finding one winner, mtx is
enqueued, which trigger a cache refresh from other processors. Reorder
mtx so that m_lock and m_qu will not be in the same cache line.
Notice that this doesn't affet arch that expose 128 bytes cache line
line aarch64 or powerpc.
Reviewed by: kib, markj
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60434
counter + 1 > 0 is a undefined behavior and nobody guarantee that it
will be honestly perform by the compiler that counter + 1 will always be
INT_MIN. We switch to INT_MAX check instead to provide a deterministic
result.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D60426
If userspace provides rate == 0, it causes kernel panic as rate is
directly used as a divident, which cannot be zero. Fix it by set the
rate to 1 if it is passed as 0.
Reviewed by: imp, emaste
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D59111
To allow normal users to render through the render device, we expose the
renderD node. This enables Wayland applications to use hardware
acceleration when running under the Linux emulator.
Additionally, libdrm and Mesa need to look up
/sys/dev/char/<major>:<minor> and <pcidev>/drm to identify the
corresponding renderer device (e.g., a renderD device). We expose this
path as well so that libdrm can locate the renderer.
Differential Revision: https://reviews.freebsd.org/D59190
AMD doesn't encode page depth in eptp. As a result, the page level is
decided by the host la57 value. Without this, it uses 4 level page and
therefore cause machine enable la57 have garbage page translation.
PR: 291768
Reviewed by: kib
Tested by: Antranig Vartanian <[email protected]>
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D57978
The previous patch assumes that we don't want to reset toggle bit in
STOPPED_STEP. However, a device can explicitly call
usbd_clear_data_toggle if necessary. As a result, instead of not
dropping the bit unconditionally, we added a field in xhci to specify
that we want to drop it, so that usbd_clear_data_toggle can handle it
correctly.
Reported by: oh
Reviewed by: kevans
Tested by: oh
Fixes: 28d85db46b ("xhci: Do not drop and add bits in xhci")
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D59186
libusb upstream uses int for register handler. This causes some library
user (like pyusb) to assume that we have int in all implementations and
therefore provides a 4 byte storage only. This causes Segmentation
fault as we will right the pointer.
Reviewed by: adrian
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D54211
(cherry picked from commit ce9ced951a)
FreeBSD's libusb has three components: libusb01, libusb10, and libusb20.
libusb20 handles communication with character devices. We now requires a
backend context for libusb20. The backend context contains contains the
capsicumized usbctrl fd and usb directory (/dev/usb) fd so that the
library user can enter the capiblity mode safely while using libusb.
libusb10 is updated to support capabilities via a context option. Since
libusb allows general read/write access, we preserve all possible
capabilities when passing backend context to libusb20. It is the
responsibility of the libusb user to call cap_enter() at an appropriate
time.
All base system tools using libusb and libusb20 have been updated to
support Capsicum.
Reviewed by: adrian, markj
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D51865
As we might change the libusb ABI in 16, we should mark thje first
version as FBSD_1.8 instead of 1.9. Since versioning patch has not
landed for a long time, it makes sense to change it directly.
Discussed with: kib
Fixes: 527a82474c ("libusb: versioning symbols")
Sponsored by: The FreeBSD Foundation
Previously, multiple frames of xfers are split into many tds. In the
refactor process, we forget to consider this. The td builder is already
allocate with enough numbers of tds. What we need to do is to fill the
normal trbs for muiltiple tds when building trbs.
PR: 297053
Tested by: phk, [email protected]
Fixes: e0b235ecd4 ("xhci: Refactor xhci_generic_setup code")
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D58465
Currently, USB request not distinguished different error and always return EIO.
However, some error are recoverable or ignorable in userspace.
Therefore, we preserve the meaning of different error to userspace then
allow userspace to decide how to use the return error.
Reviewed by: adrian
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D52244
Our USB TRB buildup subroutines were previously difficult to follow. In
setup_generic_chain_sub(), the routine filled TRB packets based on the
characteristics passed by the caller and the current state (for example,
whether the TRB was the last in the TD).
However, most TRB types (except Normal TRBs) cannot be shared across TDs.
To simplify the logic, refactor xhci_setup_generic() so that TRBs are
constructed according to their transfer type, with dedicated helper
functions for each TRB type.
Sponsored by: The FreeBSD Foundation
Assisted-by: Claude Code (Opus 4.6, Opus 4.8(1M) and Sonet 5.0)
Differential Revision: https://reviews.freebsd.org/D57130
bus_{read,write}_8 are macro wrappers around the corresponding bus_space
functions in sys/bus.h, so implementing bus_{read,write}_8 won't work.
Implement the underlying bus_space function instead.
Reviewed by: jrtc27, rlibby
Fixes: 9313f6b014
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D58301
KVM does not always use 0x40000000 as its CPUID base. For example, QEMU
adds a 0x100 offset when nested virtualization is detected and the host
exposes Hyper-V enlightenment hints. To accommodate this behavior,
switch the detection logic to use the CPUID leaf returned by do_cpuid(),
making the implementation more flexible.
See:
https://github.com/qemu/qemu/blob/master/target/i386/kvm/kvm.c#L2300
Reviewed by: kib
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D58146
The shadow buffer is addressed relative to `tg_origin`, which includes
the padding offset, whereas `gfxfb_blt` operates on coordinates without
that offset. To make `gfx_fb_copy_area` emulate the behavior of
`gfxfb_blt`, the source coordinates must include the padding offset,
while the destination coordinates must not. The original implementation
omitted the offset from the source coordinates; this change corrects
that.
Additionally, `gfx_fb_cons_display` already applies the padding offset,
so the redundant adjustment is removed.
PR: 296246
Reported by: [email protected]
Reviewed by: imp
Tested by: [email protected], [email protected],
[email protected]
Fixes: 32da2f23ae
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57821
ugen20_process() treats any non-EBUSY errno returned by USB_FS_COMPLETE
as device detach and returns LIBUSB20_ERROR_OTHER. This causes libusb10
to set device_is_gone and fail all subsequent transfer with
LIBUSB_ERROR_NO_DEVICE.
However, USB_FS_COMPLETE can also return EINVAL when a completion
references an endpoint that no longer exists, for example after
SET_INTERFACE or SET_CONFIG removes and recreates endpoints. This is a
transient condition and does not indicate device detach.
Treat EINVAL the same as EBUSY and stop draining completions. This
prevents a guest selecting an isochronous streaming altsetting from
permanently breaking the passed-through device.
Reviewed by: bapt
Event: Halifax Hackathon 202606
Location: Peggy's Cove Rock
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57542
The USB protocol defines a Zero-Length Packet (ZLP) to signal the end of
a transfer when the data size is an exact multiple of the Maximum Packet
Size (MPS). Without a ZLP in such cases, the device may not be able to
determine that the transfer has completed.
This flag is added to libusb to allow the user send a ZLP in the end
of libusb_xfer.
Reviewed by: adrian
Sponsored by: The FreeBSD Foundataion
Differential Revision: https://reviews.freebsd.org/D51759
The libusb has a function to calculate the size from given interface,
alt_setting, endpoint. Implementing it by refactoring the calculating
function for libusb_get_max_iso_packet_size.
Reviewed by: adrian
Sponsored By: FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D51225
Implement libusb_set_option for API compatibility of libusb upstream
The implementation status of each option is as following:
LIBUSB_OPTION_LOG_LEVEL: just like libusb_set_debug
LIBUSB_OPTION_LOG_CB: add callback support for DPRINTF
LIBUSB_OPTION_NO_DEVICE_DISCOVERY: disable initialization of devd and
netlink when register. Also, create no thread when registration of
callback happens.
LIBUSB_OPTION_USE_USBDK: no needed as USBDK is for Windows
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D50818
libusb provide functions to parse interface association descriptor. This
descriptor indicates that a function is composed by multiple interface
and which interfaces is associate to the target function. This
descriptor is not a separate USB require, instead, it comes with the
config descriptor.
Reviewed by: adrian
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D50958
Previously, gfxfb_blt flushed the framebuffer on every call. Since a
single drawing operation may invoke gfxfb_blt multiple times, this can
result in unnecessary flushes.
Instead, write updates to the shadow buffer (when present) and mark the
affected area as dirty. Flushing is deferred so multiple gfxfb_blt calls
can be coalesced into a single update. As before, only the dirty region
is flushed.
This fixes the slow bootloader problem in some platforms.
Reviewed by: imp, adrian, obiwac
Tested by: obiwac, jrm
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57373
FreeBSD's USB framework supports detaching kernel drivers to allow
user space applications to exclusively claim USB interfaces. However,
it lacked support for reattaching the kernel driver afterward.
This commit adds the missing functionality, enabling user space
to return control of the device back to the kernel.
Reviewed by: lwhsu
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D52122
C++ exposes cfenv functions via using ::func. Our name-mangling
mechanism rewrites all function calls causing symbols such as
std::feclearexcept to be transformed into std::__feclearexcept_int.
Since no such function exists, compilation fails.
The using ::feclearexpect declarations themselves are unaffected because
they are not function calls, which further exposes the mismatch
As a result, enable the fast path only for C and fall back to the slow
path in C++.
Reviewed by: kib
Fixes: 5bc64b7d41
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57450
In package control mode, the performance of all cores depends on the
most recent value written to the request field. If the last write comes
from an E-core, all cores are forced to align with the E-core
performance level, resulting in significant performance degradation.
Therefore, package control is disabled on hybrid-core systems.
Reviewed by: olce
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Sponsored by: Framework Computer Inc
Differential Revision: https://reviews.freebsd.org/D57377
This id is for SPI flash instead of spi bus
Fixes: 39e297bf54 ("ig4iic: Add PantherLake IDs")
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Sponsored by: Framework Computer Inc
Accessing usb_xfer_queue requires bus lock, we added this missing lock
in here to prevent racing issue.
Reviewed by: adrian
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57293
Drop and Add bits reset the data toggle for high-speed devices in XHCI.
The toggle bit represents the sequence number in USB 2.0 transfers. However,
a device can only recognize that the toggle bit has been reset while in
the HALT state. As a result, the host and device toggle values may
become mismatched, causing xHCI to reject the packet. This issue was
observed while testing the EZ-USB FX2 device.
The transfer may then return to the original value after a
bi-directional TD because the toggle field is only one bit wide. This
explains the reson that we can only receive packets bi-transfer in some
case. Therefore, we do not reset the toggle bit here.
Reviewed by: adrian
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57146
chn_poll() may hold both rdch and wrch channel locks while calling
chn_trigger(rdch). chn_trigger() switches the lock order from
"channel -> dsp dev" to "dsp dev -> channel" by temporarily dropping
the channel lock before acquiring the dsp lock.
However, only rdch was unlocked during the transition while wrch
remained locked. Since wrch is also a channel lock and witness had
already established the lock order requirement:
dsp dev -> channel
witness reports a lock order reversal when pcm_lock() is acquired while
wrch is still held.
Avoid holding rdch and wrch simultaneously during chn_trigger()
lock-order switching by only keeping the channel locks when needed.
The issue can be reliably reproduced by starting pipewire,
pipewire-pulse, and pavucontrol.
Reviewed by: christos
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57009
libusb upstream uses int for register handler. This causes some library
user (like pyusb) to assume that we have int in all implementations and
therefore provides a 4 byte storage only. This causes Segmentation
fault as we will right the pointer.
Reviewed by: adrian
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D54211
After patch 9d26b82, we don't provide recursive call protection anymore.
Therefore, to pass the test, we adjust the testcase by protecting on
caller and the testcase is to make sure the dtors is properly handled.
Reported by: siva
Reviewed by: kib
Approved by: markj (mentor)
Fixes: 9d26b82826 ("libc: Fix dtor order in __cxa_thread_atexit")
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D55893
We separate the code of CPPC and legacy pstate driver to make it easier
to read.
Reviewed by: olce
Approved by: markj (mentor)
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D55604
The thread_local variable may creates another thread_local variable
inside its dtor. This new object is immediately be registered in
__cxa_thread_atexit() and need to be freed before processing another
variable.
This fixes the libcxx test thread_local_destruction_order.pass.cpp.
Reported by: kib
Approved by: lwhsu (mentor)
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D55826
When preecise option is used, the true elapsed time should also use the
precise timer.
This fixes the test case sleep_for.signals.pass.cpp in libcxx.
Reviewed by: kib, imp
Approved by: lwhsu (mentor)
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D55824
Previously, we iterated over all CPUs using CPU_FOREACH and checked
individual bits to count valid CPUs. Refactor this to use a bitwise AND
and popcount to count the number of enabled bits directly.
Approved by: markj (mentor)
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D54474
Add parentheses between OR and AND operator to pass the compiler check.
Approved by: lwhsu (mentor)
Fixes: 3e6e4e4a0d ("hwpstate: add CPPC support for pstate driver on AMD")
MFC aftert: 2 weeks
Differential Revision: https://reviews.freebsd.org/D54465
Implement CPPC interface for AMD Pstate Driver.
This feature is only enabled when the CPUID shows it support CPPC.
The CPPC is implemneted by the following steps:
1. Write MSR to enable it.
2. Read capability registert which indicates binary value of levels
about lowest, best energy efficient, guarantee, and max performance.
3. Write request register with epp in energy balanced mode. And let
CPU and firmware to enter autonomous mode.
Also, create a sysctl handler to allow userspace to change epp value.
In intel's hwpstate, The epp value can be in package level and core level.
However, in AMD's one, there is only core level. Thus, to sync with the
intel's code, we implement package level control in software and provide
another sysctl (machdep.hwpstate_pkg_ctrl) to control it.
Reviewed by: olce, khng
Approved by: lwhsu (mentor)
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D49587
The register set information is stored as a singleton in
GetRegisterInfo_i386. However, other functions later access this
information assuming it is stored in GetSharedRegisterInfoVector. To
resolve this inconsistency, we remove the original construction logic
and instead initialize the singleton using llvm::call_once within the
appropriate function (GetSharedRegisterInfoVector_i386).
PR: 289945
Obtained from: llvm-project 41859c27842eeda1ef6ff18f3b2fb269388c0857
ShengYi Hung