Two clearly different specimens of NEC uPD720200 (one with start/stop
bug, one without) were seen to cause IOMMU faults after some Missed
Service Errors. Faulting address is immediately after a transfer ring
segment and patched dynamic debug messages revealed that the MSE was
received when waiting for a TD near the end of that segment:
[ 1.041954] xhci_hcd: Miss service interval error for slot 1 ep 2 expected TD DMA ffa08fe0
[ 1.042120] xhci_hcd: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0005 address=0xffa09000 flags=0x0000]
[ 1.042146] xhci_hcd: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0005 address=0xffa09040 flags=0x0000]
It gets even funnier if the next page is a ring segment accessible to
the HC. Below, it reports MSE in segment at ff1e8000, plows through a
zero-filled page at ff1e9000 and starts reporting events for TRBs in
page at ff1ea000 every microframe, instead of jumping to seg ff1e6000.
[ 7.041671] xhci_hcd: Miss service interval error for slot 1 ep 2 expected TD DMA ff1e8fe0
[ 7.041999] xhci_hcd: Miss service interval error for slot 1 ep 2 expected TD DMA ff1e8fe0
[ 7.042011] xhci_hcd: WARN: buffer overrun event for slot 1 ep 2 on endpoint
[ 7.042028] xhci_hcd: All TDs skipped for slot 1 ep 2. Clear skip flag.
[ 7.042134] xhci_hcd: WARN: buffer overrun event for slot 1 ep 2 on endpoint
[ 7.042138] xhci_hcd: ERROR Transfer event TRB DMA ptr not part of current TD ep_index 2 comp_code 31
[ 7.042144] xhci_hcd: Looking for event-dma 00000000ff1ea040 trb-start 00000000ff1e6820 trb-end 00000000ff1e6820
[ 7.042259] xhci_hcd: WARN: buffer overrun event for slot 1 ep 2 on endpoint
[ 7.042262] xhci_hcd: ERROR Transfer event TRB DMA ptr not part of current TD ep_index 2 comp_code 31
[ 7.042266] xhci_hcd: Looking for event-dma 00000000ff1ea050 trb-start 00000000ff1e6820 trb-end 00000000ff1e6820
At some point completion events change from Isoch Buffer Overrun to
Short Packet and the HC finally finds cycle bit mismatch in ff1ec000.
[ 7.098130] xhci_hcd: ERROR Transfer event TRB DMA ptr not part of current TD ep_index 2 comp_code 13
[ 7.098132] xhci_hcd: Looking for event-dma 00000000ff1ecc50 trb-start 00000000ff1e6820 trb-end 00000000ff1e6820
[ 7.098254] xhci_hcd: ERROR Transfer event TRB DMA ptr not part of current TD ep_index 2 comp_code 13
[ 7.098256] xhci_hcd: Looking for event-dma 00000000ff1ecc60 trb-start 00000000ff1e6820 trb-end 00000000ff1e6820
[ 7.098379] xhci_hcd: Overrun event on slot 1 ep 2
It's possible that data from the isochronous device were written to
random buffers of pending TDs on other endpoints (either IN or OUT),
other devices or even other HCs in the same IOMMU domain.
Lastly, an error from a different USB device on another HC. Was it
caused by the above? I don't know, but it may have been. The disk
was working without any other issues and generated PCIe traffic to
starve the NEC of upstream BW and trigger those MSEs. The two HCs
shared one x1 slot by means of a commercial "PCIe splitter" board.
[ 7.162604] usb 10-2: reset SuperSpeed USB device number 3 using xhci_hcd
[ 7.178990] sd 9:0:0:0: [sdb] tag#0 UNKNOWN(0x2003) Result: hostbyte=0x07 driverbyte=DRIVER_OK cmd_age=0s
[ 7.179001] sd 9:0:0:0: [sdb] tag#0 CDB: opcode=0x28 28 00 04 02 ae 00 00 02 00 00
[ 7.179004] I/O error, dev sdb, sector 67284480 op 0x0:(READ) flags 0x80700 phys_seg 5 prio class 0
Fortunately, it appears that this ridiculous bug is avoided by setting
the chain bit of Link TRBs on isochronous rings. Other ancient HCs are
known which also expect the bit to be set and they ignore Link TRBs if
it's not. Reportedly, 0.95 spec guaranteed that the bit is set.
The bandwidth-starved NEC HC running a 32KB/uframe UVC endpoint reports
tens of MSEs per second and runs into the bug within seconds. Chaining
Link TRBs allows the same workload to run for many minutes, many times.
No negative side effects seen in UVC recording and UAC playback with a
few devices at full speed, high speed and SuperSpeed.
The problem doesn't reproduce on the newer Renesas uPD720201/uPD720202
and on old Etron EJ168 and VIA VL805 (but the VL805 has other bug).
[shorten line length of log snippets in commit messge -Mathias]
Signed-off-by: Michal Pecio <michal.pecio@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-14-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ensure that an endpoint halted due to device STALL is not
restarted before a Clear_Feature(ENDPOINT_HALT) request is sent to
the device.
The host side of the endpoint may otherwise be started early by the
'Set TR Deq' command completion handler which is called if dequeue
is moved past a cancelled or halted TD.
Prevent this with a new flag set for bulk and interrupt endpoints
when a Stall Error is received. Clear it in hcd->endpoint_reset()
which is called after Clear_Feature(ENDPOINT_HALT) is sent.
Also add a debug message if a class driver queues a new URB after the
STALL. Note that class driver might not be aware of the STALL
yet when it submits the URB as URBs are given back in BH.
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-13-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Function trb_in_td() currently includes debug capabilities that are
triggered when its debug argument is set to true. The only consumer of
these debug capabilities is handle_tx_event(), which calls trb_in_td()
twice, once for its primary functionality and a second time solely for
debugging purposes if the first call returns 'NULL'.
This approach is inefficient and can lead to confusion, as trb_in_td()
executes the same code with identical arguments twice, differing only in
the debug output during the second execution.
To enhance clarity and efficiency, move the debug capabilities out of
trb_in_td() and integrates them directly into handle_tx_event().
This change reduces the argument count of trb_in_td() and ensures that
debug steps are executed only when necessary, streamlining the function's
operation.
Signed-off-by: Niklas Neronin <niklas.neronin@linux.intel.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-12-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current xHCI driver does not validate whether a page size of 4096
bytes is supported. Address the issue by setting the page size to the
value supported by the xHCI controller, as read from the Page Size
register. In the event of an unexpected value; default to a 4K page size.
Additionally, this commit removes unnecessary debug messages and instead
prints the supported and used page size once.
The xHCI controller supports page sizes of (2^{(n+12)}) bytes, where 'n'
is the Page Size Bit. Only one page size is supported, with a maximum
page size of 128 KB.
Signed-off-by: Niklas Neronin <niklas.neronin@linux.intel.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-10-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The ffs() function returns the index of the first set bit, starting from 1.
If no bits are set, it returns zero. This behavior causes an off-by-one
page size in the debug message, as the page size calculation [1]
is zero-based, while ffs() is one-based.
Fix this by subtracting one from the result of ffs(). Note that since
variable 'val' is unsigned, subtracting one from zero will result in the
maximum unsigned integer value. Consequently, the condition 'if (val < 16)'
will still function correctly.
[1], Page size: (2^(n+12)), where 'n' is the set page size bit.
Fixes: 81720ec532 ("usb: host: xhci: use ffs() in xhci_mem_init()")
Signed-off-by: Niklas Neronin <niklas.neronin@linux.intel.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-9-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If skipping is deferred to events other than Missed Service Error itsef,
it means we are running on an xHCI 1.0 host and don't know how many TDs
were missed until we reach some ordinary transfer completion event.
And in case of ring xrun, we can't know where the xrun happened either.
If we skip all pending TDs, we may prematurely give back TDs added after
the xrun had occurred, risking data loss or buffer UAF by the xHC.
If we skip none, a driver may become confused and stop working when all
its URBs are missed and appear to be "in flight" forever.
Skip exactly one TD on each xrun event - the first one that was missed,
as we can now be sure that the HC has finished processing it. Provided
that one more TD is queued before any subsequent doorbell ring, it will
become safe to skip another TD by the time we get an xrun again.
Signed-off-by: Michal Pecio <michal.pecio@gmail.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-8-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
xHCI spec rev. 1.0 allowed the TRB pointer of Missed Service events
to be NULL. Having no idea which of the queued TDs were missed and
which are waiting, we can only set a flag to skip missed TDs later.
But HCs are also allowed to give us pointer to the last missed TRB,
and this became mandatory in spec rev. 1.1 and later.
Use this pointer, if available, to immediately skip all missed TDs.
This reduces latency and risk of skipping-related bugs, because we
can now leave the skip flag cleared for future events.
Handle Missed Service Error events as 'error mid TD', if applicable,
because rev. 1.0 spec excplicitly says so in notes to 4.10.3.2 and
later revs in 4.10.3.2 and 4.11.2.5.2. Notes to 4.9.1 seem to apply.
Tested on ASM1142 and ASM3142 v1.1 xHCs which provide TRB pointers.
Tested on AMD, Etron, Renesas v1.0 xHCs which provide TRB pointers.
Tested on NEC v0.96 and VIA v1.0 xHCs which send a NULL pointer.
Change inspired by a discussion about realtime USB audio.
Link: https://lore.kernel.org/linux-usb/76e1a191-020d-4a76-97f6-237f9bd0ede0@gmx.net/T/
Signed-off-by: Michal Pecio <michal.pecio@gmail.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-7-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The TRB pointer of these events points at enqueue at the time of error
occurrence on xHCI 1.1+ HCs or it's NULL on older ones. By the time we
are handling the event, a new TD may be queued at this ring position.
I can trigger this race by rising interrupt moderation to increase IRQ
handling delay. Similar delay may occur naturally due to system load.
If this ever happens after a Missed Service Error, missed TDs will be
skipped and the new TD processed as if it matched the event. It could
be given back prematurely, risking data loss or buffer UAF by the xHC.
Don't complete TDs on xrun events and don't warn if queued TDs don't
match the event's TRB pointer, which can be NULL or a link/no-op TRB.
Don't warn if there are no queued TDs at all.
Now that it's safe, also handle xrun events if the skip flag is clear.
This ensures completion of any TD stuck in 'error mid TD' state right
before the xrun event, which could happen if a driver submits a finite
number of URBs to a buggy HC and then an error occurs on the last TD.
Signed-off-by: Michal Pecio <michal.pecio@gmail.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-6-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Up until commit d56b0b2ab1 ("usb: xhci: ensure skipped isoc TDs are
returned when isoc ring is stopped") in v6.11, the driver didn't skip
missed isochronous TDs when handling Stoppend and Stopped - Length
Invalid events. Instead, it erroneously cleared the skip flag, which
would cause the ring to get stuck, as future events won't match the
missed TD which is never removed from the queue until it's cancelled.
This buggy logic seems to have been in place substantially unchanged
since the 3.x series over 10 years ago, which probably speaks first
and foremost about relative rarity of this case in normal usage, but
by the spec I see no reason why it shouldn't be possible.
After d56b0b2ab1, TDs are immediately skipped when handling those
Stopped events. This poses a potential problem in case of Stopped -
Length Invalid, which occurs either on completed TDs (likely already
given back) or Link and No-Op TRBs. Such event won't be recognized
as matching any TD (unless it's the rare Link TRB inside a TD) and
will result in skipping all pending TDs, giving them back possibly
before they are done, risking isoc data loss and maybe UAF by HW.
As a compromise, don't skip and don't clear the skip flag on this
kind of event. Then the next event will skip missed TDs. A downside
of not handling Stopped - Length Invalid on a Link inside a TD is
that if the TD is cancelled, its actual length will not be updated
to account for TRBs (silently) completed before the TD was stopped.
I had no luck producing this sequence of completion events so there
is no compelling demonstration of any resulting disaster. It may be
a very rare, obscure condition. The sole motivation for this patch
is that if such unlikely event does occur, I'd rather risk reporting
a cancelled partially done isoc frame as empty than gamble with UAF.
This will be fixed more properly by looking at Stopped event's TRB
pointer when making skipping decisions, but such rework is unlikely
to be backported to v6.12, which will stay around for a few years.
Fixes: d56b0b2ab1 ("usb: xhci: ensure skipped isoc TDs are returned when isoc ring is stopped")
Cc: stable@vger.kernel.org
Signed-off-by: Michal Pecio <michal.pecio@gmail.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20250306144954.3507700-4-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The error `failed to get FW build information` is added for what looks
to be for misdetection of the device property firmware-name.
If the property is missing (such as on non-nvidia HW) this error shows up.
Move the error into the scope of the property parser for "firmware-name"
to avoid showing errors on systems without the firmware-name property.
Fixes: 5c9ae5a875 ("usb: typec: ucsi: ccg: add firmware flashing support")
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://lore.kernel.org/r/20250221054137.1631765-2-superm1@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The usb_sndaddr0pipe macro is only used in the hub_set_address function.
Replace it with usb_sndctrlpipe which provides the same functionality
but would also consider the endpoint device number.
If the device has not been initialised, it is safe to use
usb_sndctrlpipe in this context because udev->devnum is set to 0.
Therefore, this change does not affect behaviour, but reduces code
complexity by reusing the existing macro.
Signed-off-by: Stefan Eichenberger <stefan.eichenberger@toradex.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Link: https://lore.kernel.org/r/20250219083745.10406-1-eichest@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add support for the 'eUSB2 Isochronous Endpoint Companion Descriptor'
introduced in the recent USB 2.0 specification 'USB 2.0 Double Isochronous
IN Bandwidth' ECN.
It allows embedded USB2 (eUSB2) devices to report and use higher bandwidths
for isochronous IN transfers in order to support higher camera resolutions
on the lid of laptops and tablets with minimal change to the USB2 protocol.
The motivation for expanding USB 2.0 is further clarified in an additional
Embedded USB2 version 2.0 (eUSB2v2) supplement to the USB 2.0
specification. It points out this is optimized for performance, power and
cost by using the USB 2.0 low-voltage, power efficient PHY and half-duplex
link for the asymmetric camera bandwidth needs, avoiding the costly and
complex full-duplex USB 3.x symmetric link and gigabit receivers.
eUSB2 devices that support the higher isochronous IN bandwidth and the new
descriptor can be identified by their device descriptor bcdUSB value of
0x0220
Co-developed-by: Amardeep Rai <amardeep.rai@intel.com>
Signed-off-by: Amardeep Rai <amardeep.rai@intel.com>
Signed-off-by: Kannappan R <r.kannappan@intel.com>
Co-developed-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Link: https://lore.kernel.org/r/20250220141339.1939448-1-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According to the dt-bindings there are some platforms, which have a
dedicated USB power domain for DWC2 IP core supply. If the power domain
is switched off during system suspend then all USB register will lose
their settings.
Use GUSBCFG_TOUTCAL as a canary to detect that the power domain has
been powered off during suspend. Since the GOTGCTL_CURMODE_HOST doesn't
match on all platform with the current mode, additionally backup
GINTSTS. This works reliable to decide which registers should be
restored.
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20250217134132.36786-4-wahrenst@gmx.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The DWC2 runtime PM code reuses similar patterns to backup and
restore the registers. So consolidate them in USB mode specific
variants. This also has the advantage it is reusable for further
PM improvements.
Special care is taken for DCFG register during device mode restore.
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Link: https://lore.kernel.org/r/20250217134132.36786-3-wahrenst@gmx.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pull Kbuild fixes from Masahiro Yamada:
- Fix annoying logs when building tools in parallel
- Fix the Debian linux-headers package build again
- Fix the target triple detection for userspace programs on Clang
* tag 'kbuild-fixes-v6.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
modpost: Fix a few typos in a comment
kbuild: userprogs: fix bitsize and target detection on clang
kbuild: fix linux-headers package build when $(CC) cannot link userspace
tools: fix annoying "mkdir -p ..." logs when building tools in parallel
Michal Pecio
Mathias Nyman
Niklas Neronin
Kaustabh Chakraborty
Kever Yang
Jonathan Neuschäfer
Mario Limonciello
Xu Yang
Rob Herring (Arm)
Dr. David Alan Gilbert
AngeloGioacchino Del Regno
Stefan Eichenberger
Kannappan R
Igor Belwon
Johan Hovold
Stefan Wahren
Suraj Patil
Krzysztof Kozlowski
Catalin Popescu
Greg Kroah-Hartman
Linus Torvalds