xhci: Fix repeated xhci wake after suspend due to uncleared internal wake state

commit d26c00e7276fc92b18c253d69e872f6b03832bad upstream.

If port terminations are detected in suspend, but link never reaches U0
then xHCI may have an internal uncleared wake state that will cause an
immediate wake after suspend.

This wake state is normally cleared when driver clears the PORT_CSC bit,
which is set after a device is enabled and in U0.

Write 1 to clear PORT_CSC for ports that don't have anything connected
when suspending. This makes sure any pending internal wake states in
xHCI are cleared.

Cc: stable@vger.kernel.org
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20210311115353.2137560-5-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: xhci: Fix ASMedia ASM1042A and ASM3242 DMA addressing

commit b71c669ad8390dd1c866298319ff89fe68b45653 upstream.

I've confirmed that both the ASMedia ASM1042A and ASM3242 have the same
problem as the ASM1142 and ASM2142/ASM3142, where they lose some of the
upper bits of 64-bit DMA addresses. As with the other chips, this can
cause problems on systems where the upper bits matter, and adding the
XHCI_NO_64BIT_SUPPORT quirk completely fixes the issue.

Cc: stable@vger.kernel.org
Signed-off-by: Forest Crossman <cyrozap@gmail.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20210311115353.2137560-4-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xhci: Improve detection of device initiated wake signal.

commit 253f588c70f66184b1f3a9bbb428b49bbda73e80 upstream.

A xHC USB 3 port might miss the first wake signal from a USB 3 device
if the port LFPS reveiver isn't enabled fast enough after xHC resume.

xHC host will anyway be resumed by a PME# signal, but will go back to
suspend if no port activity is seen.
The device resends the U3 LFPS wake signal after a 100ms delay, but
by then host is already suspended, starting all over from the
beginning of this issue.

USB 3 specs say U3 wake LFPS signal is sent for max 10ms, then device
needs to delay 100ms before resending the wake.

Don't suspend immediately if port activity isn't detected in resume.
Instead add a retry. If there is no port activity then delay for 120ms,
and re-check for port activity.

Cc: <stable@vger.kernel.org>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20210311115353.2137560-3-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: xhci: do not perform Soft Retry for some xHCI hosts

commit a4a251f8c23518899d2078c320cf9ce2fa459c9f upstream.

On some systems rt2800usb and mt7601u devices are unable to operate since
commit f8f80be501aa ("xhci: Use soft retry to recover faster from
transaction errors")

Seems that some xHCI controllers can not perform Soft Retry correctly,
affecting those devices.

To avoid the problem add xhci->quirks flag that restore pre soft retry
xhci behaviour for affected xHCI controllers. Currently those are
AMD_PROMONTORYA_4 and AMD_PROMONTORYA_2, since it was confirmed
by the users: on those xHCI hosts issue happen and is gone after
disabling Soft Retry.

[minor commit message rewording for checkpatch -Mathias]

Fixes: f8f80be501aa ("xhci: Use soft retry to recover faster from transaction errors")
Cc: <stable@vger.kernel.org> # 4.20+
Reported-by: Bernhard <bernhard.gebetsberger@gmx.at>
Tested-by: Bernhard <bernhard.gebetsberger@gmx.at>
Signed-off-by: Stanislaw Gruszka <stf_xl@wp.pl>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=202541
Link: https://lore.kernel.org/r/20210311115353.2137560-2-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: renesas_usbhs: Clear PIPECFG for re-enabling pipe with other EPNUM

commit b1d25e6ee57c2605845595b6c61340d734253eb3 upstream.

According to the datasheet, this controller has a restriction
which "set an endpoint number so that combinations of the DIR bit and
the EPNUM bits do not overlap.". However, since the udc core driver is
possible to assign a bulk pipe as an interrupt endpoint, an endpoint
number may not match the pipe number. After that, when user rebinds
another gadget driver, this driver broke the restriction because
the driver didn't clear any configuration in usb_ep_disable().

Example:
 # modprobe g_ncm
 Then, EP3 = pipe 3, EP4 = pipe 4, EP5 = pipe 6
 # rmmod g_ncm
 # modprobe g_hid
 Then, EP3 = pipe 6, EP4 = pipe 7.
 So, pipe 3 and pipe 6 are set as EP3.

So, clear PIPECFG register in usbhs_pipe_free().

Fixes: dfb87b8bfe09 ("usb: renesas_usbhs: gadget: fix re-enabling pipe without re-connecting")
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/1615168538-26101-1-git-send-email-yoshihiro.shimoda.uh@renesas.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: usblp: fix a hang in poll() if disconnected

commit 9de2c43acf37a17dc4c69ff78bb099b80fb74325 upstream.

Apparently an application that opens a device and calls select()
on it, will hang if the decice is disconnected. It's a little
surprising that we had this bug for 15 years, but apparently
nobody ever uses select() with a printer: only write() and read(),
and those work fine. Well, you can also select() with a timeout.

The fix is modeled after devio.c. A few other drivers check the
condition first, then do not add the wait queue in case the
device is disconnected. We doubt that's completely race-free.
So, this patch adds the process first, then locks properly
and checks for the disconnect.

Reviewed-by: Zqiang <qiang.zhang@windriver.com>
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210303221053.1cf3313e@suzdal.zaitcev.lan
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: dwc3: qcom: Honor wakeup enabled/disabled state

commit 2664deb0930643149d61cddbb66ada527ae180bd upstream.

The dwc3-qcom currently enables wakeup interrupts unconditionally
when suspending, however this should not be done when wakeup is
disabled (e.g. through the sysfs attribute power/wakeup). Only
enable wakeup interrupts when device_may_wakeup() returns true.

Fixes: a4333c3a6ba9 ("usb: dwc3: Add Qualcomm DWC3 glue driver")
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210302103659.v2.1.I44954d9e1169f2cf5c44e6454d357c75ddfa99a2@changeid
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: dwc3: qcom: Add missing DWC3 OF node refcount decrement

commit 1cffb1c66499a9db9a735473778abf8427d16287 upstream.

of_get_child_by_name() increments the reference counter of the OF node it
managed to find. So after the code is done using the device node, the
refcount must be decremented. Add missing of_node_put() invocation then
to the dwc3_qcom_of_register_core() method, since DWC3 OF node is being
used only there.

Fixes: a4333c3a6ba9 ("usb: dwc3: Add Qualcomm DWC3 glue driver")
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Link: https://lore.kernel.org/r/20210212205521.14280-1-Sergey.Semin@baikalelectronics.ru
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: gadget: f_uac1: stop playback on function disable

commit cc2ac63d4cf72104e0e7f58bb846121f0f51bb19 upstream.

There is missing playback stop/cleanup in case of
gadget's ->disable callback that happens on
events like USB host resetting or gadget disconnection

Fixes: 0591bc236015 ("usb: gadget: add f_uac1 variant based on a new u_audio api")
Cc: <stable@vger.kernel.org> # 4.13+
Signed-off-by: Ruslan Bilovol <ruslan.bilovol@gmail.com>
Link: https://lore.kernel.org/r/1614599375-8803-3-git-send-email-ruslan.bilovol@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: gadget: f_uac2: always increase endpoint max_packet_size by one audio slot

commit 789ea77310f0200c84002884ffd628e2baf3ad8a upstream.

As per UAC2 Audio Data Formats spec (2.3.1.1 USB Packets),
if the sampling rate is a constant, the allowable variation
of number of audio slots per virtual frame is +/- 1 audio slot.

It means that endpoint should be able to accept/send +1 audio
slot.

Previous endpoint max_packet_size calculation code
was adding sometimes +1 audio slot due to DIV_ROUND_UP
behaviour which was rounding up to closest integer.
However this doesn't work if the numbers are divisible.

It had no any impact with Linux hosts which ignore
this issue, but in case of more strict Windows it
caused rejected enumeration

Thus always add +1 audio slot to endpoint's max packet size

Fixes: 913e4a90b6f9 ("usb: gadget: f_uac2: finalize wMaxPacketSize according to bandwidth")
Cc: Peter Chen <peter.chen@freescale.com>
Cc: <stable@vger.kernel.org> #v4.3+
Signed-off-by: Ruslan Bilovol <ruslan.bilovol@gmail.com>
Link: https://lore.kernel.org/r/1614599375-8803-2-git-send-email-ruslan.bilovol@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: gadget: u_ether: Fix a configfs return code

commit 650bf52208d804ad5ee449c58102f8dc43175573 upstream.

If the string is invalid, this should return -EINVAL instead of 0.

Fixes: 73517cf49bd4 ("usb: gadget: add RNDIS configfs options for class/subclass/protocol")
Cc: stable <stable@vger.kernel.org>
Acked-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/YCqZ3P53yyIg5cn7@mwanda
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Goodix Fingerprint device is not a modem

commit 4d8654e81db7346f915eca9f1aff18f385cab621 upstream.

The CDC ACM driver is false matching the Goodix Fingerprint device
against the USB_CDC_ACM_PROTO_AT_V25TER.

The Goodix Fingerprint device is a biometrics sensor that should be
handled in user-space. libfprint has some support for Goodix
fingerprint sensors, although not for this particular one. It is
possible that the vendor allocates a PID per OEM (Lenovo, Dell etc).
If this happens to be the case then more devices from the same vendor
could potentially match the ACM modem module table.

Signed-off-by: Yorick de Wid <ydewid@gmail.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210213144901.53199-1-ydewid@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: cqhci: Fix random crash when remove mmc module/card

commit f06391c45e83f9a731045deb23df7cc3814fd795 upstream.

[ 6684.493350] Unable to handle kernel paging request at virtual address ffff800011c5b0f0
[ 6684.498531] mmc0: card 0001 removed
[ 6684.501556] Mem abort info:
[ 6684.509681]   ESR = 0x96000047
[ 6684.512786]   EC = 0x25: DABT (current EL), IL = 32 bits
[ 6684.518394]   SET = 0, FnV = 0
[ 6684.521707]   EA = 0, S1PTW = 0
[ 6684.524998] Data abort info:
[ 6684.528236]   ISV = 0, ISS = 0x00000047
[ 6684.532986]   CM = 0, WnR = 1
[ 6684.536129] swapper pgtable: 4k pages, 48-bit VAs, pgdp=0000000081b22000
[ 6684.543923] [ffff800011c5b0f0] pgd=00000000bffff003, p4d=00000000bffff003, pud=00000000bfffe003, pmd=00000000900e1003, pte=0000000000000000
[ 6684.557915] Internal error: Oops: 96000047 [#1] PREEMPT SMP
[ 6684.564240] Modules linked in: sdhci_esdhc_imx(-) sdhci_pltfm sdhci cqhci mmc_block mmc_core fsl_jr_uio caam_jr caamkeyblob_desc caamhash_desc caamalg_desc crypto_engine rng_core authenc libdes crct10dif_ce flexcan can_dev caam error [last unloaded: mmc_core]
[ 6684.587281] CPU: 0 PID: 79138 Comm: kworker/0:3H Not tainted 5.10.9-01410-g3ba33182767b-dirty #10
[ 6684.596160] Hardware name: Freescale i.MX8DXL EVK (DT)
[ 6684.601320] Workqueue: kblockd blk_mq_run_work_fn

[ 6684.606094] pstate: 40000005 (nZcv daif -PAN -UAO -TCO BTYPE=--)
[ 6684.612286] pc : cqhci_request+0x148/0x4e8 [cqhci]
^GMessage from syslogd@  at Thu Jan  1 01:51:24 1970 ...[ 6684.617085] lr : cqhci_request+0x314/0x4e8 [cqhci]
[ 6684.626734] sp : ffff80001243b9f0
[ 6684.630049] x29: ffff80001243b9f0 x28: ffff00002c3dd000
[ 6684.635367] x27: 0000000000000001 x26: 0000000000000001
[ 6684.640690] x25: ffff00002c451000 x24: 000000000000000f
[ 6684.646007] x23: ffff000017e71c80 x22: ffff00002c451000
[ 6684.651326] x21: ffff00002c0f3550 x20: ffff00002c0f3550
[ 6684.656651] x19: ffff000017d46880 x18: ffff00002cea1500
[ 6684.661977] x17: 0000000000000000 x16: 0000000000000000
[ 6684.667294] x15: 000001ee628e3ed1 x14: 0000000000000278
[ 6684.672610] x13: 0000000000000001 x12: 0000000000000001
[ 6684.677927] x11: 0000000000000000 x10: 0000000000000000
[ 6684.683243] x9 : 000000000000002b x8 : 0000000000001000
[ 6684.688560] x7 : 0000000000000010 x6 : ffff00002c0f3678
[ 6684.693886] x5 : 000000000000000f x4 : ffff800011c5b000
[ 6684.699211] x3 : 000000000002d988 x2 : 0000000000000008
[ 6684.704537] x1 : 00000000000000f0 x0 : 0002d9880008102f
[ 6684.709854] Call trace:
[ 6684.712313]  cqhci_request+0x148/0x4e8 [cqhci]
[ 6684.716803]  mmc_cqe_start_req+0x58/0x68 [mmc_core]
[ 6684.721698]  mmc_blk_mq_issue_rq+0x460/0x810 [mmc_block]
[ 6684.727018]  mmc_mq_queue_rq+0x118/0x2b0 [mmc_block]

The problem occurs when cqhci_request() get called after cqhci_disable() as
it leads to access of allocated memory that has already been freed. Let's
fix the problem by calling cqhci_disable() a bit later in the remove path.

Signed-off-by: Frank Li <Frank.Li@nxp.com>
Diagnosed-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20210303174248.542175-1-Frank.Li@nxp.com
Fixes: f690f4409ddd ("mmc: mmc: Enable CQE's")
Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: core: Fix partition switch time for eMMC

commit 66fbacccbab91e6e55d9c8f1fc0910a8eb6c81f7 upstream.

Avoid the following warning by always defining partition switch time:

 [    3.209874] mmc1: unspecified timeout for CMD6 - use generic
 [    3.222780] ------------[ cut here ]------------
 [    3.233363] WARNING: CPU: 1 PID: 111 at drivers/mmc/core/mmc_ops.c:575 __mmc_switch+0x200/0x204

Reported-by: Paul Fertser <fercerpav@gmail.com>
Fixes: 1c447116d017 ("mmc: mmc: Fix partition switch timeout for some eMMCs")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/168bbfd6-0c5b-5ace-ab41-402e7937c46e@intel.com
Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

software node: Fix node registration

commit 8891123f9cbb9c1ee531e5a87fa116f0af685c48 upstream.

Software node can not be registered before its parent.

Fixes: 80488a6b1d3c ("software node: Add support for static node descriptors")
Cc: 5.10+ <stable@vger.kernel.org> # 5.10+
Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Tested-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/dasd: fix hanging IO request during DASD driver unbind

commit 66f669a272898feb1c69b770e1504aa2ec7723d1 upstream.

Prevent that an IO request is build during device shutdown initiated by
a driver unbind. This request will never be able to be processed or
canceled and will hang forever. This will lead also to a hanging unbind.

Fix by checking not only if the device is in READY state but also check
that there is no device offline initiated before building a new IO request.

Fixes: e443343e509a ("s390/dasd: blk-mq conversion")
Cc: <stable@vger.kernel.org> # v4.14+
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Tested-by: Bjoern Walk <bwalk@linux.ibm.com>
Reviewed-by: Jan Hoeppner <hoeppner@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/dasd: fix hanging DASD driver unbind

commit 7d365bd0bff3c0310c39ebaffc9a8458e036d666 upstream.

In case of an unbind of the DASD device driver the function
dasd_generic_remove() is called which shuts down the device.
Among others this functions removes the int_handler from the cdev.
During shutdown the device cancels all outstanding IO requests and waits
for completion of the clear request.
Unfortunately the clear interrupt will never be received when there is no
interrupt handler connected.

Fix by moving the int_handler removal after the call to the state machine
where no request or interrupt is outstanding.

Cc: stable@vger.kernel.org
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Tested-by: Bjoern Walk <bwalk@linux.ibm.com>
Reviewed-by: Jan Hoeppner <hoeppner@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64: kasan: fix page_alloc tagging with DEBUG_VIRTUAL

commit 86c83365ab76e4b43cedd3ce07a07d32a4dc79ba upstream.

When CONFIG_DEBUG_VIRTUAL is enabled, the default page_to_virt() macro
implementation from include/linux/mm.h is used. That definition doesn't
account for KASAN tags, which leads to no tags on page_alloc allocations.

Provide an arm64-specific definition for page_to_virt() when
CONFIG_DEBUG_VIRTUAL is enabled that takes care of KASAN tags.

Fixes: 2813b9c02962 ("kasan, mm, arm64: tag non slab memory allocated via pagealloc")
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/4b55b35202706223d3118230701c6a59749d9b72.1615219501.git.andreyknvl@google.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert 95ebabde382c ("capabilities: Don't allow writing ambiguous v3 file capabilities")

commit 3b0c2d3eaa83da259d7726192cf55a137769012f upstream.

It turns out that there are in fact userspace implementations that
care and this recent change caused a regression.

https://github.com/containers/buildah/issues/3071

As the motivation for the original change was future development,
and the impact is existing real world code just revert this change
and allow the ambiguity in v3 file caps.

Cc: stable@vger.kernel.org
Fixes: 95ebabde382c ("capabilities: Don't allow writing ambiguous v3 file capabilities")
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: usb-audio: Apply the control quirk to Plantronics headsets

commit 06abcb18b3a021ba1a3f2020cbefb3ed04e59e72 upstream.

Other Plantronics headset models seem requiring the same workaround as
C320-M to add the 20ms delay for the control messages, too.  Apply the
workaround generically for devices with the vendor ID 0x047f.

Note that the problem didn't surface before 5.11 just with luck.
Since 5.11 got a big code rewrite about the stream handling, the
parameter setup procedure has changed, and this seemed triggering the
problem more often.

BugLink: https://bugzilla.suse.com/show_bug.cgi?id=1182552
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210304085009.4770-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: usb-audio: Fix "cannot get freq eq" errors on Dell AE515 sound bar

commit fec60c3bc5d1713db2727cdffc638d48f9c07dc3 upstream.

Dell AE515 sound bar (413c:a506) spews the error messages when the
driver tries to read the current sample frequency, hence it needs to
be on the list in snd_usb_get_sample_rate_quirk().

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=211551
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210304083021.2152-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda: Avoid spurious unsol event handling during S3/S4

commit 5ff9dde42e8c72ed8102eb8cb62e03f9dc2103ab upstream.

When HD-audio bus receives unsolicited events during its system
suspend/resume (S3 and S4) phase, the controller driver may still try
to process events although the codec chips are already (or yet)
powered down.  This might screw up the codec communication, resulting
in CORB/RIRB errors.  Such events should be rather skipped, as the
codec chip status such as the jack status will be fully refreshed at
the system resume time.

Since we're tracking the system suspend/resume state in codec
power.power_state field, let's add the check in the common unsol event
handler entry point to filter out such events.

BugLink: https://bugzilla.suse.com/show_bug.cgi?id=1182377
Tested-by: Abhishek Sahu <abhsahu@nvidia.com>
Cc: <stable@vger.kernel.org> # 183ab39eb0ea: ALSA: hda: Initialize power_state
Link: https://lore.kernel.org/r/20210310112809.9215-3-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda: Flush pending unsolicited events before suspend

commit 13661fc48461282e43fe8f76bf5bf449b3d40687 upstream.

The HD-audio controller driver processes the unsolicited events via
its work asynchronously, and this might be pending when the system
goes to suspend.  When a lengthy event handling like ELD byte reads is
running, this might trigger unexpected accesses among suspend/resume
procedure, typically seen with Nvidia driver that still requires the
handling via unsolicited event verbs for ELD updates.

This patch adds the flush of unsol_work to assure that pending events
are processed before going into suspend.

Buglink: https://bugzilla.suse.com/show_bug.cgi?id=1182377
Reported-and-tested-by: Abhishek Sahu <abhsahu@nvidia.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210310112809.9215-2-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda: Drop the BATCH workaround for AMD controllers

commit 28e96c1693ec1cdc963807611f8b5ad400431e82 upstream.

The commit c02f77d32d2c ("ALSA: hda - Workaround for crackled sound on
AMD controller (1022:1457)") introduced a few workarounds for the
recent AMD HD-audio controller, and one of them is the forced BATCH
PCM mode so that PulseAudio avoids the timer-based scheduling.  This
was thought to cover for some badly working applications, but this
actually worsens for more others.  In total, this wasn't a good idea
to enforce it.

This is a partial revert of the commit above for dropping the PCM
BATCH enforcement part to recover from the regression again.

Fixes: c02f77d32d2c ("ALSA: hda - Workaround for crackled sound on AMD controller (1022:1457)")
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=195303
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210308160726.22930-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/ca0132: Add Sound BlasterX AE-5 Plus support

commit f15c5c11abfbf8909eb30598315ecbec2311cfdc upstream.

The new AE-5 Plus model has a different Subsystem ID compared to the
non-plus model. Adding the new id to the list of quirks.

Signed-off-by: Simeon Simeonoff <sim.simeonoff@gmail.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/998cafbe10b648f724ee33570553f2d780a38963.camel@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/hdmi: Cancel pending works before suspend

commit eea46a0879bcca23e15071f9968c0f6e6596e470 upstream.

The per_pin->work might be still floating at the suspend, and this may
hit the access to the hardware at an unexpected timing.  Cancel the
work properly at the suspend callback for avoiding the buggy access.

Note that the bug doesn't trigger easily in the recent kernels since
the work is queued only when the repoll count is set, and usually it's
only at the resume callback, but it's still possible to hit in
theory.

BugLink: https://bugzilla.suse.com/show_bug.cgi?id=1182377
Reported-and-tested-by: Abhishek Sahu <abhsahu@nvidia.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210310112809.9215-4-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: usb: Add Plantronics C320-M USB ctrl msg delay quirk

commit fc7c5c208eb7bc2df3a9f4234f14eca250001cb6 upstream.

The microphone in the Plantronics C320-M headset will randomly
fail to initialize properly, at least when using Microsoft Teams.
Introducing a 20ms delay on the control messages appears to
resolve the issue.

Link: https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/issues/1065
Tested-by: Andreas Kempe <kempe@lysator.liu.se>
Signed-off-by: John Ernberg <john.ernberg@actia.se>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210303181405.39835-1-john.ernberg@actia.se
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: target: core: Prevent underflow for service actions

[ Upstream commit 14d24e2cc77411301e906a8cf41884739de192de ]

TCM buffer length doesn't necessarily equal 8 + ADDITIONAL LENGTH which
might be considered an underflow in case of Data-In size being greater than
8 + ADDITIONAL LENGTH. So truncate buffer length to prevent underflow.

Link: https://lore.kernel.org/r/20210209072202.41154-3-a.miloserdov@yadro.com
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Bodo Stroesser <bostroesser@gmail.com>
Signed-off-by: Aleksandr Miloserdov <a.miloserdov@yadro.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: target: core: Add cmd length set before cmd complete

[ Upstream commit 1c73e0c5e54d5f7d77f422a10b03ebe61eaed5ad ]

TCM doesn't properly handle underflow case for service actions. One way to
prevent it is to always complete command with
target_complete_cmd_with_length(), however it requires access to data_sg,
which is not always available.

This change introduces target_set_cmd_data_length() function which allows
to set command data length before completing it.

Link: https://lore.kernel.org/r/20210209072202.41154-2-a.miloserdov@yadro.com
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Bodo Stroesser <bostroesser@gmail.com>
Signed-off-by: Aleksandr Miloserdov <a.miloserdov@yadro.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: libiscsi: Fix iscsi_prep_scsi_cmd_pdu() error handling

[ Upstream commit d28d48c699779973ab9a3bd0e5acfa112bd4fdef ]

If iscsi_prep_scsi_cmd_pdu() fails we try to add it back to the cmdqueue,
but we leave it partially setup. We don't have functions that can undo the
pdu and init task setup. We only have cleanup_task which can clean up both
parts. So this has us just fail the cmd and go through the standard cleanup
routine and then have the SCSI midlayer retry it like is done when it fails
in the queuecommand path.

Link: https://lore.kernel.org/r/20210207044608.27585-2-michael.christie@oracle.com
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sysctl.c: fix underflow value setting risk in vm_table

[ Upstream commit 3b3376f222e3ab58367d9dd405cafd09d5e37b7c ]

Apart from subsystem specific .proc_handler handler, all ctl_tables with
extra1 and extra2 members set should use proc_dointvec_minmax instead of
proc_dointvec, or the limit set in extra* never work and potentially echo
underflow values(negative numbers) is likely make system unstable.

Especially vfs_cache_pressure and zone_reclaim_mode, -1 is apparently not
a valid value, but we can set to them.  And then kernel may crash.

# echo -1 > /proc/sys/vm/vfs_cache_pressure

Link: https://lkml.kernel.org/r/20201223105535.2875-1-linf@wangsu.com
Signed-off-by: Lin Feng <linf@wangsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/smp: __smp_rescan_cpus() - move cpumask away from stack

[ Upstream commit 62c8dca9e194326802b43c60763f856d782b225c ]

Avoid a potentially large stack frame and overflow by making
"cpumask_t avail" a static variable. There is no concurrent
access due to the existing locking.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i40e: Fix memory leak in i40e_probe

[ Upstream commit 58cab46c622d6324e47bd1c533693c94498e4172 ]

Struct i40e_veb is allocated in function i40e_setup_pf_switch, and
stored to an array field veb inside struct i40e_pf. However when
i40e_setup_misc_vector fails, this memory leaks.

Fix this by calling exit and teardown functions.

Signed-off-by: Keita Suzuki <keitasuzuki.park@sslab.ics.keio.ac.jp>
Tested-by: Tony Brelinski <tonyx.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

PCI: Fix pci_register_io_range() memory leak

[ Upstream commit f6bda644fa3a7070621c3bf12cd657f69a42f170 ]

Kmemleak reports:

  unreferenced object 0xc328de40 (size 64):
    comm "kworker/1:1", pid 21, jiffies 4294938212 (age 1484.670s)
    hex dump (first 32 bytes):
      00 00 00 00 00 00 00 00 e0 d8 fc eb 00 00 00 00  ................
      00 00 10 fe 00 00 00 00 00 00 00 00 00 00 00 00  ................

  backtrace:
    [<ad758d10>] pci_register_io_range+0x3c/0x80
    [<2c7f139e>] of_pci_range_to_resource+0x48/0xc0
    [<f079ecc8>] devm_of_pci_get_host_bridge_resources.constprop.0+0x2ac/0x3ac
    [<e999753b>] devm_of_pci_bridge_init+0x60/0x1b8
    [<a895b229>] devm_pci_alloc_host_bridge+0x54/0x64
    [<e451ddb0>] rcar_pcie_probe+0x2c/0x644

In case a PCI host driver's probe is deferred, the same I/O range may be
allocated again, and be ignored, causing a memory leak.

Fix this by (a) letting logic_pio_register_range() return -EEXIST if the
passed range already exists, so pci_register_io_range() will free it, and
by (b) making pci_register_io_range() not consider -EEXIST an error
condition.

Link: https://lore.kernel.org/r/20210202100332.829047-1-geert+renesas@glider.be
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

kbuild: clamp SUBLEVEL to 255

[ Upstream commit 9b82f13e7ef316cdc0a8858f1349f4defce3f9e0 ]

Right now if SUBLEVEL becomes larger than 255 it will overflow into the
territory of PATCHLEVEL, causing havoc in userspace that tests for
specific kernel version.

While userspace code tests for MAJOR and PATCHLEVEL, it doesn't test
SUBLEVEL at any point as ABI changes don't happen in the context of
stable tree.

Thus, to avoid overflows, simply clamp SUBLEVEL to it's maximum value in
the context of LINUX_VERSION_CODE. This does not affect "make
kernelversion" and such.

Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

PCI: mediatek: Add missing of_node_put() to fix reference leak

[ Upstream commit 42814c438aac79746d310f413a27d5b0b959c5de ]

The for_each_available_child_of_node helper internally makes use of the
of_get_next_available_child() which performs an of_node_get() on each
iteration when searching for next available child node.

Should an available child node be found, then it would return a device
node pointer with reference count incremented, thus early return from
the middle of the loop requires an explicit of_node_put() to prevent
reference count leak.

To stop the reference leak, explicitly call of_node_put() before
returning after an error occurred.

Link: https://lore.kernel.org/r/20210120184810.3068794-1-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

PCI: xgene-msi: Fix race in installing chained irq handler

[ Upstream commit a93c00e5f975f23592895b7e83f35de2d36b7633 ]

Fix a race where a pending interrupt could be received and the handler
called before the handler's data has been setup, by converting to
irq_set_chained_handler_and_data().

See also 2cf5a03cb29d ("PCI/keystone: Fix race in installing chained IRQ
handler").

Based on the mail discussion, it seems ok to drop the error handling.

Link: https://lore.kernel.org/r/20210115212435.19940-3-martin@kaiser.cx
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Input: applespi - don't wait for responses to commands indefinitely.

[ Upstream commit 0ce1ac23149c6da939a5926c098c270c58c317a0 ]

The response to a command may never arrive or it may be corrupted (and
hence dropped) for some reason. While exceedingly rare, when it did
happen it blocked all further commands. One way to fix this was to
do a suspend/resume. However, recovering automatically seems like a
nicer option. Hence this puts a time limit (1 sec) on how long we're
willing to wait for a response, after which we assume it got lost.

Signed-off-by: Ronald Tschalär <ronald@innovation.ch>
Link: https://lore.kernel.org/r/20210217190718.11035-1-ronald@innovation.ch
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sparc64: Use arch_validate_flags() to validate ADI flag

[ Upstream commit 147d8622f2a26ef34beacc60e1ed8b66c2fa457f ]

When userspace calls mprotect() to enable ADI on an address range,
do_mprotect_pkey() calls arch_validate_prot() to validate new
protection flags. arch_validate_prot() for sparc looks at the first
VMA associated with address range to verify if ADI can indeed be
enabled on this address range. This has two issues - (1) Address
range might cover multiple VMAs while arch_validate_prot() looks at
only the first VMA, (2) arch_validate_prot() peeks at VMA without
holding mmap lock which can result in race condition.

arch_validate_flags() from commit c462ac288f2c ("mm: Introduce
arch_validate_flags()") allows for VMA flags to be validated for all
VMAs that cover the address range given by user while holding mmap
lock. This patch updates sparc code to move the VMA check from
arch_validate_prot() to arch_validate_flags() to fix above two
issues.

Suggested-by: Jann Horn <jannh@google.com>
Suggested-by: Christoph Hellwig <hch@infradead.org>
Suggested-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sparc32: Limit memblock allocation to low memory

[ Upstream commit bda166930c37604ffa93f2425426af6921ec575a ]

Commit cca079ef8ac29a7c02192d2bad2ffe4c0c5ffdd0 changed sparc32 to use
memblocks instead of bootmem, but also made high memory available via
memblock allocation which does not work together with e.g. phys_to_virt
and can lead to kernel panic.

This changes back to only low memory being allocatable in the early
stages, now using memblock allocation.

Signed-off-by: Andreas Larsson <andreas@gaisler.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu/amd: Fix performance counter initialization

[ Upstream commit 6778ff5b21bd8e78c8bd547fd66437cf2657fd9b ]

Certain AMD platforms enable power gating feature for IOMMU PMC,
which prevents the IOMMU driver from updating the counter while
trying to validate the PMC functionality in the init_iommu_perf_ctr().
This results in disabling PMC support and the following error message:

    "AMD-Vi: Unable to read/write to IOMMU perf counter"

To workaround this issue, disable power gating temporarily by programming
the counter source to non-zero value while validating the counter,
and restore the prior state afterward.

Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Tested-by: Tj (Elloe Linux) <ml.linux@elloe.vision>
Link: https://lore.kernel.org/r/20210208122712.5048-1-suravee.suthikulpanit@amd.com
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=201753
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/64: Fix stack trace not displaying final frame

[ Upstream commit e3de1e291fa58a1ab0f471a4b458eff2514e4b5f ]

In commit bf13718bc57a ("powerpc: show registers when unwinding
interrupt frames") we changed our stack dumping logic to show the full
registers whenever we find an interrupt frame on the stack.

However we didn't notice that on 64-bit this doesn't show the final
frame, ie. the interrupt that brought us in from userspace, whereas on
32-bit it does.

That is due to confusion about the size of that last frame. The code
in show_stack() calls validate_sp(), passing it STACK_INT_FRAME_SIZE
to check the sp is at least that far below the top of the stack.

However on 64-bit that size is too large for the final frame, because
it includes the red zone, but we don't allocate a red zone for the
first frame.

So add a new define that encodes the correct size for 32-bit and
64-bit, and use it in show_stack().

This results in the full trace being shown on 64-bit, eg:

  sysrq: Trigger a crash
  Kernel panic - not syncing: sysrq triggered crash
  CPU: 0 PID: 83 Comm: sh Not tainted 5.11.0-rc2-gcc-8.2.0-00188-g571abcb96b10-dirty #649
  Call Trace:
  [c00000000a1c3ac0] [c000000000897b70] dump_stack+0xc4/0x114 (unreliable)
  [c00000000a1c3b00] [c00000000014334c] panic+0x178/0x41c
  [c00000000a1c3ba0] [c00000000094e600] sysrq_handle_crash+0x40/0x50
  [c00000000a1c3c00] [c00000000094ef98] __handle_sysrq+0xd8/0x210
  [c00000000a1c3ca0] [c00000000094f820] write_sysrq_trigger+0x100/0x188
  [c00000000a1c3ce0] [c0000000005559dc] proc_reg_write+0x10c/0x1b0
  [c00000000a1c3d10] [c000000000479950] vfs_write+0xf0/0x360
  [c00000000a1c3d60] [c000000000479d9c] ksys_write+0x7c/0x140
  [c00000000a1c3db0] [c00000000002bf5c] system_call_exception+0x19c/0x2c0
  [c00000000a1c3e10] [c00000000000d35c] system_call_common+0xec/0x278
  --- interrupt: c00 at 0x7fff9fbab428
  NIP:  00007fff9fbab428 LR: 000000001000b724 CTR: 0000000000000000
  REGS: c00000000a1c3e80 TRAP: 0c00   Not tainted  (5.11.0-rc2-gcc-8.2.0-00188-g571abcb96b10-dirty)
  MSR:  900000000280f033 <SF,HV,VEC,VSX,EE,PR,FP,ME,IR,DR,RI,LE>  CR: 22002884  XER: 00000000
  IRQMASK: 0
  GPR00: 0000000000000004 00007fffc3cb8960 00007fff9fc59900 0000000000000001
  GPR04: 000000002a4b32d0 0000000000000002 0000000000000063 0000000000000063
  GPR08: 000000002a4b32d0 0000000000000000 0000000000000000 0000000000000000
  GPR12: 0000000000000000 00007fff9fcca9a0 0000000000000000 0000000000000000
  GPR16: 0000000000000000 0000000000000000 0000000000000000 00000000100b8fd0
  GPR20: 000000002a4b3485 00000000100b8f90 0000000000000000 0000000000000000
  GPR24: 000000002a4b0440 00000000100e77b8 0000000000000020 000000002a4b32d0
  GPR28: 0000000000000001 0000000000000002 000000002a4b32d0 0000000000000001
  NIP [00007fff9fbab428] 0x7fff9fbab428
  LR [000000001000b724] 0x1000b724
  --- interrupt: c00

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210209141627.2898485-1-mpe@ellerman.id.au
Signed-off-by: Sasha Levin <sashal@kernel.org>

HID: logitech-dj: add support for the new lightspeed connection iteration

[ Upstream commit fab3a95654eea01d6b0204995be8b7492a00d001 ]

This new connection type is the new iteration of the Lightspeed
connection and will probably be used in some of the newer gaming
devices. It is currently use in the G Pro X Superlight.

This patch should be backported to older versions, as currently the
driver will panic when seing the unsupported connection. This isn't
an issue when using the receiver that came with the device, as Logitech
has been using different PIDs when they change the connection type, but
is an issue when using a generic receiver (well, generic Lightspeed
receiver), which is the case of the one in the Powerplay mat. Currently,
the only generic Ligthspeed receiver we support, and the only one that
exists AFAIK, is ther Powerplay.

As it stands, the driver will panic when seeing a G Pro X Superlight
connected to the Powerplay receiver and won't send any input events to
userspace! The kernel will warn about this so the issue should be easy
to identify, but it is still very worrying how hard it will fail :(

[915977.398471] logitech-djreceiver 0003:046D:C53A.0107: unusable device of type UNKNOWN (0x0f) connected on slot 1

Signed-off-by: Filipe Laíns <lains@riseup.net>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/perf: Record counter overflow always if SAMPLE_IP is unset

[ Upstream commit d137845c973147a22622cc76c7b0bc16f6206323 ]

While sampling for marked events, currently we record the sample only
if the SIAR valid bit of Sampled Instruction Event Register (SIER) is
set. SIAR_VALID bit is used for fetching the instruction address from
Sampled Instruction Address Register(SIAR). But there are some
usecases, where the user is interested only in the PMU stats at each
counter overflow and the exact IP of the overflow event is not
required. Dropping SIAR invalid samples will fail to record some of
the counter overflows in such cases.

Example of such usecase is dumping the PMU stats (event counts) after
some regular amount of instructions/events from the userspace (ex: via
ptrace). Here counter overflow is indicated to userspace via signal
handler, and captured by monitoring and enabling I/O signaling on the
event file descriptor. In these cases, we expect to get
sample/overflow indication after each specified sample_period.

Perf event attribute will not have PERF_SAMPLE_IP set in the
sample_type if exact IP of the overflow event is not requested. So
while profiling if SAMPLE_IP is not set, just record the counter
overflow irrespective of SIAR_VALID check.

Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
[mpe: Reflow comment and if formatting]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1612516492-1428-1-git-send-email-atrajeev@linux.vnet.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc: improve handling of unrecoverable system reset

[ Upstream commit 11cb0a25f71818ca7ab4856548ecfd83c169aa4d ]

If an unrecoverable system reset hits in process context, the system
does not have to panic. Similar to machine check, call nmi_exit()
before die().

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210130130852.2952424-26-npiggin@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: stm32: make spurious and overrun interrupts visible

[ Upstream commit c64e7efe46b7de21937ef4b3594d9b1fc74f07df ]

We do not expect to receive spurious interrupts so rise a warning
if it happens.

RX overrun is an error condition that signals a corrupted RX
stream both in dma and in irq modes. Report the error and
abort the transfer in either cases.

Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Link: https://lore.kernel.org/r/1612551572-495-9-git-send-email-alain.volmat@foss.st.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/pci: Add ppc_md.discover_phbs()

[ Upstream commit 5537fcb319d016ce387f818dd774179bc03217f5 ]

On many powerpc platforms the discovery and initalisation of
pci_controllers (PHBs) happens inside of setup_arch(). This is very early
in boot (pre-initcalls) and means that we're initialising the PHB long
before many basic kernel services (slab allocator, debugfs, a real ioremap)
are available.

On PowerNV this causes an additional problem since we map the PHB registers
with ioremap(). As of commit d538aadc2718 ("powerpc/ioremap: warn on early
use of ioremap()") a warning is printed because we're using the "incorrect"
API to setup and MMIO mapping in searly boot. The kernel does provide
early_ioremap(), but that is not intended to create long-lived MMIO
mappings and a seperate warning is printed by generic code if
early_ioremap() mappings are "leaked."

This is all fixable with dumb hacks like using early_ioremap() to setup
the initial mapping then replacing it with a real ioremap later on in
boot, but it does raise the question: Why the hell are we setting up the
PHB's this early in boot?

The old and wise claim it's due to "hysterical rasins." Aside from amused
grapes there doesn't appear to be any real reason to maintain the current
behaviour. Already most of the newer embedded platforms perform PHB
discovery in an arch_initcall and between the end of setup_arch() and the
start of initcalls none of the generic kernel code does anything PCI
related. On powerpc scanning PHBs occurs in a subsys_initcall so it should
be possible to move the PHB discovery to a core, postcore or arch initcall.

This patch adds the ppc_md.discover_phbs hook and a core_initcall stub that
calls it. The core_initcalls are the earliest to be called so this will
any possibly issues with dependency between initcalls. This isn't just an
academic issue either since on pseries and PowerNV EEH init occurs in an
arch_initcall and depends on the pci_controllers being available, similarly
the creation of pci_dns occurs at core_initcall_sync (i.e. between core and
postcore initcalls). These problems need to be addressed seperately.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
[mpe: Make discover_phbs() static]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20201103043523.916109-1-oohall@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

Platform: OLPC: Fix probe error handling

[ Upstream commit cec551ea0d41c679ed11d758e1a386e20285b29d ]

Reset ec_priv if probe ends unsuccessfully.

Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Link: https://lore.kernel.org/r/20210126073740.10232-2-lkundrak@v3.sk
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mmc: mediatek: fix race condition between msdc_request_timeout and irq

[ Upstream commit 0354ca6edd464a2cf332f390581977b8699ed081 ]

when get request SW timeout, if CMD/DAT xfer done irq coming right now,
then there is race between the msdc_request_timeout work and irq handler,
and the host->cmd and host->data may set to NULL in irq handler. also,
current flow ensure that only one path can go to msdc_request_done(), so
no need check the return value of cancel_delayed_work().

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Link: https://lore.kernel.org/r/20201218071611.12276-1-chaotian.jing@mediatek.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mmc: mxs-mmc: Fix a resource leak in an error handling path in 'mxs_mmc_probe()'

[ Upstream commit 0bb7e560f821c7770973a94e346654c4bdccd42c ]

If 'mmc_of_parse()' fails, we must undo the previous 'dma_request_chan()'
call.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/20201208203527.49262-1-christophe.jaillet@wanadoo.fr
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

udf: fix silent AED tagLocation corruption

[ Upstream commit 63c9e47a1642fc817654a1bc18a6ec4bbcc0f056 ]

When extending a file, udf_do_extend_file() may enter following empty
indirect extent. At the end of udf_do_extend_file() we revert prev_epos
to point to the last written extent. However if we end up not adding any
further extent in udf_do_extend_file(), the reverting points prev_epos
into the header area of the AED and following updates of the extents
(in udf_update_extents()) will corrupt the header.

Make sure that we do not follow indirect extent if we are not going to
add any more extents so that returning back to the last written extent
works correctly.

Link: https://lore.kernel.org/r/20210107234116.6190-2-magnani@ieee.org
Signed-off-by: Steven J. Magnani <magnani@ieee.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i2c: rcar: optimize cacheline to minimize HW race condition

[ Upstream commit 25c2e0fb5fefb8d7847214cf114d94c7aad8e9ce ]

'flags' and 'io' are needed first, so they should be at the beginning of
the private struct.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i2c: rcar: faster irq code to minimize HW race condition

[ Upstream commit c7b514ec979e23a08c411f3d8ed39c7922751422 ]

To avoid the HW race condition on R-Car Gen2 and earlier, we need to
write to ICMCR as soon as possible in the interrupt handler. We can
improve this by writing a static value instead of masking out bits.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: phy: fix save wrong speed and duplex problem if autoneg is on

[ Upstream commit d9032dba5a2b2bbf0fdce67c8795300ec9923b43 ]

If phy uses generic driver and autoneg is on, enter command
"ethtool -s eth0 speed 50" will not change phy speed actually, but
command "ethtool eth0" shows speed is 50Mb/s because phydev->speed
has been set to 50 and no update later.

And duplex setting has same problem too.

However, if autoneg is on, phy only changes speed and duplex according to
phydev->advertising, but not phydev->speed and phydev->duplex. So in this
case, phydev->speed and phydev->duplex don't need to be set in function
phy_ethtool_ksettings_set() if autoneg is on.

Fixes: 51e2a3846eab ("PHY: Avoid unnecessary aneg restarts")
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: enetc: initialize RFS/RSS memories for unused ports too

[ Upstream commit 3222b5b613db558e9a494bbf53f3c984d90f71ea ]

Michael reports that since linux-next-20210211, the AER messages for ECC
errors have started reappearing, and this time they can be reliably
reproduced with the first ping on one of his LS1028A boards.

$ ping 1[   33.258069] pcieport 0000:00:1f.0: AER: Multiple Corrected error received: 0000:00:00.0
72.16.0.1
PING [   33.267050] pcieport 0000:00:1f.0: AER: can't find device of ID0000
172.16.0.1 (172.16.0.1): 56 data bytes
64 bytes from 172.16.0.1: seq=0 ttl=64 time=17.124 ms
64 bytes from 172.16.0.1: seq=1 ttl=64 time=0.273 ms

$ devmem 0x1f8010e10 32
0xC0000006

It isn't clear why this is necessary, but it seems that for the errors
to go away, we must clear the entire RFS and RSS memory, not just for
the ports in use.

Sadly the code is structured in such a way that we can't have unified
logic for the used and unused ports. For the minimal initialization of
an unused port, we need just to enable and ioremap the PF memory space,
and a control buffer descriptor ring. Unused ports must then free the
CBDR because the driver will exit, but used ports can not pick up from
where that code path left, since the CBDR API does not reinitialize a
ring when setting it up, so its producer and consumer indices are out of
sync between the software and hardware state. So a separate
enetc_init_unused_port function was created, and it gets called right
after the PF memory space is enabled.

Fixes: 07bf34a50e32 ("net: enetc: initialize the RFS and RSS memories")
Reported-by: Michael Walle <michael@walle.cc>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: Michael Walle <michael@walle.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: hns3: fix error mask definition of flow director

[ Upstream commit ae85ddda0f1b341b2d25f5a5e0eff1d42b6ef3df ]

Currently, some bit filed definitions of flow director TCAM
configuration command are incorrect. Since the wrong MSB is
always 0, and these fields are assgined in order, so it still works.

Fix it by redefine them.

Fixes: 117328680288 ("net: hns3: Add input key and action config support for flow director")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: rc: compile rc-cec.c into rc-core

commit f09f9f93afad770a04b35235a0aa465fcc8d6e3d upstream.

The rc-cec keymap is unusual in that it can't be built as a module,
instead it is registered directly in rc-main.c if CONFIG_MEDIA_CEC_RC
is set. This is because it can be called from drm_dp_cec_set_edid() via
cec_register_adapter() in an asynchronous context, and it is not
allowed to use request_module() to load rc-cec.ko in that case. Trying to
do so results in a 'WARN_ON_ONCE(wait && current_is_async())'.

Since this keymap is only used if CONFIG_MEDIA_CEC_RC is set, we
just compile this keymap into the rc-core module and never as a
separate module.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Fixes: 2c6d1fffa1d9 (drm: add support for DisplayPort CEC-Tunneling-over-AUX)
Reported-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: v4l: vsp1: Fix bru null pointer access

commit ac8d82f586c8692b501cb974604a71ef0e22a04c upstream.

RZ/G2L SoC has only BRS. This patch fixes null pointer access,when only
BRS is enabled.

Fixes: cbb7fa49c7466("media: v4l: vsp1: Rename BRU to BRx")
Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: v4l: vsp1: Fix uif null pointer access

commit 6732f313938027a910e1f7351951ff52c0329e70 upstream.

RZ/G2L SoC has no UIF. This patch fixes null pointer access, when UIF
module is not used.

Fixes: 5e824f989e6e8("media: v4l: vsp1: Integrate DISCOM in display pipeline")
Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: usbtv: Fix deadlock on suspend

commit 8a7e27fd5cd696ba564a3f62cedef7269cfd0723 upstream.

usbtv doesn't support power management, so on system suspend the
.disconnect callback of the driver is called. The teardown sequence
includes a call to snd_card_free. Its implementation waits until the
refcount of the sound card device drops to zero, however, if its file is
open, snd_card_file_add takes a reference, which can't be dropped during
the suspend, because the userspace processes are already frozen at this
point. snd_card_free waits for completion forever, leading to a hang on
suspend.

This commit fixes this deadlock condition by replacing snd_card_free
with snd_card_free_when_closed, that doesn't wait until all references
are released, allowing suspend to progress.

Fixes: 63ddf68de52e ("[media] usbtv: add audio support")
Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sh_eth: fix TRSCER mask for R7S9210

commit 165bc5a4f30eee4735845aa7dbd6b738643f2603 upstream.

According  to the RZ/A2M Group User's Manual: Hardware, Rev. 2.00,
the TRSCER register has bit 9 reserved, hence we can't use the driver's
default TRSCER mask.  Add the explicit initializer for sh_eth_cpu_data::
trscer_err_mask for R7S9210.

Fixes: 6e0bb04d0e4f ("sh_eth: Add R7S9210 support")
Signed-off-by: Sergey Shtylyov <s.shtylyov@omprussia.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

qxl: Fix uninitialised struct field head.surface_id

commit 738acd49eb018feb873e0fac8f9517493f6ce2c7 upstream.

The surface_id struct field in head is not being initialized and
static analysis warns that this is being passed through to
dev->monitors_config->heads[i] on an assignment. Clear up this
warning by initializing it to zero.

Addresses-Coverity: ("Uninitialized scalar variable")
Fixes: a6d3c4d79822 ("qxl: hook monitors_config updates into crtc, not encoder.")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20210304094928.2280722-1-colin.king@canonical.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/crypto: return -EFAULT if copy_to_user() fails

commit 942df4be7ab40195e2a839e9de81951a5862bc5b upstream.

The copy_to_user() function returns the number of bytes remaining to be
copied, but we want to return -EFAULT if the copy doesn't complete.

Fixes: e06670c5fe3b ("s390: vfio-ap: implement VFIO_DEVICE_GET_INFO ioctl")
Signed-off-by: Wang Qing <wangqing@vivo.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Link: https://lore.kernel.org/r/1614600502-16714-1-git-send-email-wangqing@vivo.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/cio: return -EFAULT if copy_to_user() fails

commit d9c48a948d29bcb22f4fe61a81b718ef6de561a0 upstream.

Fixes: 120e214e504f ("vfio: ccw: realize VFIO_DEVICE_G(S)ET_IRQ_INFO ioctls")
Signed-off-by: Eric Farman <farman@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm: meson_drv add shutdown function

commit fa0c16caf3d73ab4d2e5d6fa2ef2394dbec91791 upstream.

Problem: random stucks on reboot stage about 1/20 stuck/reboots
// debug kernel log
[    4.496660] reboot: kernel restart prepare CMD:(null)
[    4.498114] meson_ee_pwrc c883c000.system-controller:power-controller: shutdown begin
[    4.503949] meson_ee_pwrc c883c000.system-controller:power-controller: shutdown domain 0:VPU...
...STUCK...

Solution: add shutdown function to meson_drm driver
// debug kernel log
[    5.231896] reboot: kernel restart prepare CMD:(null)
[    5.246135] [drm:meson_drv_shutdown]
...
[    5.259271] meson_ee_pwrc c883c000.system-controller:power-controller: shutdown begin
[    5.274688] meson_ee_pwrc c883c000.system-controller:power-controller: shutdown domain 0:VPU...
[    5.338331] reboot: Restarting system
[    5.358293] psci: PSCI_0_2_FN_SYSTEM_RESET reboot_mode:0 cmd:(null)
bl31 reboot reason: 0xd
bl31 reboot reason: 0x0
system cmd  1.
...REBOOT...

Tested: on VIM1 VIM2 VIM3 VIM3L khadas sbcs - 1000+ successful reboots
and Odroid boards, WeTek Play2 (GXBB)

Fixes: bbbe775ec5b5 ("drm: Add support for Amlogic Meson Graphic Controller")
Signed-off-by: Artem Lapkin <art@khadas.com>
Tested-by: Christian Hewitt <christianshewitt@gmail.com>
Acked-by: Neil Armstrong <narmstrong@baylibre.com>
Acked-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210302042202.3728113-1-art@khadas.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/shmem-helper: Don't remove the offset in vm_area_struct pgoff

commit 11d5a4745e00e73745774671dbf2fb07bd6e2363 upstream.

When mmapping the shmem, it would previously adjust the pgoff in the
vm_area_struct to remove the fake offset that is added to be able to
identify the buffer. This patch removes the adjustment and makes the
fault handler use the vm_fault address to calculate the page offset
instead. Although using this address is apparently discouraged, several
DRM drivers seem to be doing it anyway.

The problem with removing the pgoff is that it prevents
drm_vma_node_unmap from working because that searches the mapping tree
by address. That doesn't work because all of the mappings are at offset
0. drm_vma_node_unmap is being used by the shmem helpers when purging
the buffer.

This fixes a bug in Panfrost which is using drm_gem_shmem_purge. Without
this the mapping for the purged buffer can still be accessed which might
mean it would access random pages from other buffers

v2: Don't check whether the unsigned page_offset is less than 0.

Cc: stable@vger.kernel.org
Fixes: 17acb9f35ed7 ("drm/shmem: Add madvise state and purge helpers")
Signed-off-by: Neil Roberts <nroberts@igalia.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210223155125.199577-3-nroberts@igalia.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/shmem-helper: Check for purged buffers in fault handler

commit d611b4a0907cece060699f2fd347c492451cd2aa upstream.

When a buffer is madvised as not needed and then purged, any attempts to
access the buffer from user-space should cause a bus fault. This patch
adds a check for that.

Cc: stable@vger.kernel.org
Fixes: 17acb9f35ed7 ("drm/shmem: Add madvise state and purge helpers")
Signed-off-by: Neil Roberts <nroberts@igalia.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210223155125.199577-2-nroberts@igalia.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/compat: Clear bounce structures

commit de066e116306baf3a6a62691ac63cfc0b1dabddb upstream.

Some of them have gaps, or fields we don't clear. Native ioctl code
does full copies plus zero-extends on size mismatch, so nothing can
leak. But compat is more hand-rolled so need to be careful.

None of these matter for performance, so just memset.

Also I didn't fix up the CONFIG_DRM_LEGACY or CONFIG_DRM_AGP ioctl, those
are security holes anyway.

Acked-by: Maxime Ripard <mripard@kernel.org>
Reported-by: syzbot+620cf21140fc7e772a5d@syzkaller.appspotmail.com # vblank ioctl
Cc: syzbot+620cf21140fc7e772a5d@syzkaller.appspotmail.com
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210222100643.400935-1-daniel.vetter@ffwll.ch
(cherry picked from commit e926c474ebee404441c838d18224cd6f246a71b7)
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bnxt_en: reliably allocate IRQ table on reset to avoid crash

commit 20d7d1c5c9b11e9f538ed4a2289be106de970d3e upstream.

The following trace excerpt corresponds with a NULL pointer dereference
of 'bp->irq_tbl' in bnxt_setup_inta() on an Aarch64 system after many
device resets:

    Unable to handle kernel NULL pointer dereference at ... 000000d
    ...
    pc : string+0x3c/0x80
    lr : vsnprintf+0x294/0x7e0
    sp : ffff00000f61ba70 pstate : 20000145
    x29: ffff00000f61ba70 x28: 000000000000000d
    x27: ffff0000009c8b5a x26: ffff00000f61bb80
    x25: ffff0000009c8b5a x24: 0000000000000012
    x23: 00000000ffffffe0 x22: ffff000008990428
    x21: ffff00000f61bb80 x20: 000000000000000d
    x19: 000000000000001f x18: 0000000000000000
    x17: 0000000000000000 x16: ffff800b6d0fb400
    x15: 0000000000000000 x14: ffff800b7fe31ae8
    x13: 00001ed16472c920 x12: ffff000008c6b1c9
    x11: ffff000008cf0580 x10: ffff00000f61bb80
    x9 : 00000000ffffffd8 x8 : 000000000000000c
    x7 : ffff800b684b8000 x6 : 0000000000000000
    x5 : 0000000000000065 x4 : 0000000000000001
    x3 : ffff0a00ffffff04 x2 : 000000000000001f
    x1 : 0000000000000000 x0 : 000000000000000d
    Call trace:
    string+0x3c/0x80
    vsnprintf+0x294/0x7e0
    snprintf+0x44/0x50
    __bnxt_open_nic+0x34c/0x928 [bnxt_en]
    bnxt_open+0xe8/0x238 [bnxt_en]
    __dev_open+0xbc/0x130
    __dev_change_flags+0x12c/0x168
    dev_change_flags+0x20/0x60
    ...

Ordinarily, a call to bnxt_setup_inta() (not in trace due to inlining)
would not be expected on a system supporting MSIX at all. However, if
bnxt_init_int_mode() does not end up being called after the call to
bnxt_clear_int_mode() in bnxt_fw_reset_close(), then the driver will
think that only INTA is supported and bp->irq_tbl will be NULL,
causing the above crash.

In the error recovery scenario, we call bnxt_clear_int_mode() in
bnxt_fw_reset_close() early in the sequence. Ordinarily, we will
call bnxt_init_int_mode() in bnxt_hwrm_if_change() after we
reestablish communication with the firmware after reset.  However,
if the sequence has to abort before we call bnxt_init_int_mode() and
if the user later attempts to re-open the device, then it will cause
the crash above.

We fix it in 2 ways:

1. Check for bp->irq_tbl in bnxt_setup_int_mode(). If it is NULL, call
bnxt_init_init_mode().

2. If we need to abort in bnxt_hwrm_if_change() and cannot complete
the error recovery sequence, set the BNXT_STATE_ABORT_ERR flag.  This
will cause more drastic recovery at the next attempt to re-open the
device, including a call to bnxt_init_int_mode().

Fixes: 3bc7d4a352ef ("bnxt_en: Add BNXT_STATE_IN_FW_RESET state.")
Reviewed-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/cio: return -EFAULT if copy_to_user() fails again

commit 51c44babdc19aaf882e1213325a0ba291573308f upstream.

The copy_to_user() function returns the number of bytes remaining to be
copied, but we want to return -EFAULT if the copy doesn't complete.

Fixes: e01bcdd61320 ("vfio: ccw: realize VFIO_DEVICE_GET_REGION_INFO ioctl")
Signed-off-by: Wang Qing <wangqing@vivo.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Link: https://lore.kernel.org/r/1614600093-13992-1-git-send-email-wangqing@vivo.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: hns3: fix bug when calculating the TCAM table info

commit b36fc875bcdee56865c444a2cdae17d354a6d5f5 upstream.

The function hclge_fd_convert_tuple() is used to convert tuples
and tuples mask to TCAM x and y.  But it misuses the source mac
as source mac mask when convert INNER_SRC_MAC, which may cause
the flow director rule works unexpectedly. So fix it.

Fixes: 117328680288 ("net: hns3: Add input key and action config support for flow director")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: hns3: fix query vlan mask value error for flow director

commit c75ec148a316e8cf52274d16b9b422703b96f5ce upstream.

Currently, the driver returns VLAN_VID_MASK for vlan mask field,
when get flow director rule information for rule doesn't use vlan.
It may cause the vlan mask value display as 0xf000 in this
case, like below:

estuary:/$ ethtool -u eth1
50 RX rings available
Total 1 rules

Filter: 2
Rule Type: TCP over IPv4
Src IP addr: 0.0.0.0 mask: 255.255.255.255
Dest IP addr: 0.0.0.0 mask: 255.255.255.255
TOS: 0x0 mask: 0xff
Src port: 0 mask: 0xffff
Dest port: 0 mask: 0xffff
VLAN EtherType: 0x0 mask: 0xffff
VLAN: 0x0 mask: 0xf000
User-defined: 0x1234 mask: 0x0
Action: Direct to queue 3

Fix it by return 0.

Fixes: 05c2314fe6a8 ("net: hns3: Add support for rule query of flow director")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf traceevent: Ensure read cmdlines are null terminated.

commit 137a5258939aca56558f3a23eb229b9c4b293917 upstream.

Issue detected by address sanitizer.

Fixes: cd4ceb63438e9e28 ("perf util: Save pid-cmdline mapping into tracing header")
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20210226221431.1985458-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests: forwarding: Fix race condition in mirror installation

commit edcbf5137f093b5502f5f6b97cce3cbadbde27aa upstream.

When mirroring to a gretap in hardware the device expects to be
programmed with the egress port and all the encapsulating headers. This
requires the driver to resolve the path the packet will take in the
software data path and program the device accordingly.

If the path cannot be resolved (in this case because of an unresolved
neighbor), then mirror installation fails until the path is resolved.
This results in a race that causes the test to sometimes fail.

Fix this by setting the neighbor's state to permanent, so that it is
always valid.

Fixes: b5b029399fa6d ("selftests: forwarding: mirror_gre_bridge_1d_vlan: Add STP test")
Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: stmmac: fix watchdog timeout during suspend/resume stress test

commit c511819d138de38e1637eedb645c207e09680d0f upstream.

stmmac_xmit() call stmmac_tx_timer_arm() at the end to modify tx timer to
do the transmission cleanup work. Imagine such a situation, stmmac enters
suspend immediately after tx timer modified, it's expire callback
stmmac_tx_clean() would not be invoked. This could affect BQL, since
netdev_tx_sent_queue() has been called, but netdev_tx_completed_queue()
have not been involved, as a result, dql_avail(&dev_queue->dql) finally
always return a negative value.

__dev_queue_xmit->__dev_xmit_skb->qdisc_run->__qdisc_run->qdisc_restart->dequeue_skb:
if ((q->flags & TCQ_F_ONETXQUEUE) &&
netif_xmit_frozen_or_stopped(txq)) // __QUEUE_STATE_STACK_XOFF is set

Net core will stop transmitting any more. Finillay, net watchdong would timeout.
To fix this issue, we should call netdev_tx_reset_queue() in stmmac_resume().

Fixes: 54139cf3bb33 ("net: stmmac: adding multiple buffers for rx")
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: stmmac: stop each tx channel independently

commit a3e860a83397bf761ec1128a3f0ba186445992c6 upstream.

If clear GMAC_CONFIG_TE bit, it would stop all tx channels, but users
may only want to stop specific tx channel.

Fixes: 48863ce5940f ("stmmac: add DMA support for GMAC 4.xx")
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ixgbe: fail to create xfrm offload of IPsec tunnel mode SA

commit d785e1fec60179f534fbe8d006c890e5ad186e51 upstream.

Based on talks and indirect references ixgbe IPsec offlod do not
support IPsec tunnel mode offload. It can only support IPsec transport
mode offload. Now explicitly fail when creating non transport mode SA
with offload to avoid false performance expectations.

Fixes: 63a67fe229ea ("ixgbe: add ipsec offload add and remove SA")
Signed-off-by: Antony Antony <antony@phenome.org>
Acked-by: Shannon Nelson <snelson@pensando.io>
Tested-by: Tony Brelinski <tonyx.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: qrtr: fix error return code of qrtr_sendmsg()

commit 179d0ba0c454057a65929c46af0d6ad986754781 upstream.

When sock_alloc_send_skb() returns NULL to skb, no error return code of
qrtr_sendmsg() is assigned.
To fix this bug, rc is assigned with -ENOMEM in this case.

Fixes: 194ccc88297a ("net: qrtr: Support decoding incoming v2 packets")
Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: davicom: Fix regulator not turned off on driver removal

commit cf9e60aa69ae6c40d3e3e4c94dd6c8de31674e9b upstream.

We must disable the regulator that was enabled in the probe function.

Fixes: 7994fe55a4a2 ("dm9000: Add regulator and reset support to dm9000")
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: davicom: Fix regulator not turned off on failed probe

commit ac88c531a5b38877eba2365a3f28f0c8b513dc33 upstream.

When the probe fails or requests to be defered, we must disable the
regulator that was previously enabled.

Fixes: 7994fe55a4a2 ("dm9000: Add regulator and reset support to dm9000")
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: lapbether: Remove netif_start_queue / netif_stop_queue

commit f7d9d4854519fdf4d45c70a4d953438cd88e7e58 upstream.

For the devices in this driver, the default qdisc is "noqueue",
because their "tx_queue_len" is 0.

In function "__dev_queue_xmit" in "net/core/dev.c", devices with the
"noqueue" qdisc are specially handled. Packets are transmitted without
being queued after a "dev->flags & IFF_UP" check. However, it's possible
that even if this check succeeds, "ops->ndo_stop" may still have already
been called. This is because in "__dev_close_many", "ops->ndo_stop" is
called before clearing the "IFF_UP" flag.

If we call "netif_stop_queue" in "ops->ndo_stop", then it's possible in
"__dev_queue_xmit", it sees the "IFF_UP" flag is present, and then it
checks "netif_xmit_stopped" and finds that the queue is already stopped.
In this case, it will complain that:
"Virtual device ... asks to queue packet!"

To prevent "__dev_queue_xmit" from generating this complaint, we should
not call "netif_stop_queue" in "ops->ndo_stop".

We also don't need to call "netif_start_queue" in "ops->ndo_open",
because after a netdev is allocated and registered, the
"__QUEUE_STATE_DRV_XOFF" flag is initially not set, so there is no need
to call "netif_start_queue" to clear it.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xie He <xie.he.0141@gmail.com>
Acked-by: Martin Schiller <ms@dev.tdt.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cipso,calipso: resolve a number of problems with the DOI refcounts

commit ad5d07f4a9cd671233ae20983848874731102c08 upstream.

The current CIPSO and CALIPSO refcounting scheme for the DOI
definitions is a bit flawed in that we:

1. Don't correctly match gets/puts in netlbl_cipsov4_list().
2. Decrement the refcount on each attempt to remove the DOI from the
   DOI list, only removing it from the list once the refcount drops
   to zero.

This patch fixes these problems by adding the missing "puts" to
netlbl_cipsov4_list() and introduces a more conventional, i.e.
not-buggy, refcounting mechanism to the DOI definitions.  Upon the
addition of a DOI to the DOI list, it is initialized with a refcount
of one, removing a DOI from the list removes it from the list and
drops the refcount by one; "gets" and "puts" behave as expected with
respect to refcounts, increasing and decreasing the DOI's refcount by
one.

Fixes: b1edeb102397 ("netlabel: Replace protocol/NetLabel linking with refrerence counts")
Fixes: d7cce01504a0 ("netlabel: Add support for removing a CALIPSO DOI.")
Reported-by: syzbot+9ec037722d2603a9f52e@syzkaller.appspotmail.com
Signed-off-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

netdevsim: init u64 stats for 32bit hardware

commit 863a42b289c22df63db62b10fc2c2ffc237e2125 upstream.

Init the u64 stats in order to avoid the lockdep prints on the 32bit
hardware like

 INFO: trying to register non-static key.
 the code is fine but needs lockdep annotation.
 turning off the locking correctness validator.
 CPU: 0 PID: 4695 Comm: syz-executor.0 Not tainted 5.11.0-rc5-syzkaller #0
 Hardware name: ARM-Versatile Express
 Backtrace:
 [<826fc5b8>] (dump_backtrace) from [<826fc82c>] (show_stack+0x18/0x1c arch/arm/kernel/traps.c:252)
 [<826fc814>] (show_stack) from [<8270d1f8>] (__dump_stack lib/dump_stack.c:79 [inline])
 [<826fc814>] (show_stack) from [<8270d1f8>] (dump_stack+0xa8/0xc8 lib/dump_stack.c:120)
 [<8270d150>] (dump_stack) from [<802bf9c0>] (assign_lock_key kernel/locking/lockdep.c:935 [inline])
 [<8270d150>] (dump_stack) from [<802bf9c0>] (register_lock_class+0xabc/0xb68 kernel/locking/lockdep.c:1247)
 [<802bef04>] (register_lock_class) from [<802baa2c>] (__lock_acquire+0x84/0x32d4 kernel/locking/lockdep.c:4711)
 [<802ba9a8>] (__lock_acquire) from [<802be840>] (lock_acquire.part.0+0xf0/0x554 kernel/locking/lockdep.c:5442)
 [<802be750>] (lock_acquire.part.0) from [<802bed10>] (lock_acquire+0x6c/0x74 kernel/locking/lockdep.c:5415)
 [<802beca4>] (lock_acquire) from [<81560548>] (seqcount_lockdep_reader_access include/linux/seqlock.h:103 [inline])
 [<802beca4>] (lock_acquire) from [<81560548>] (__u64_stats_fetch_begin include/linux/u64_stats_sync.h:164 [inline])
 [<802beca4>] (lock_acquire) from [<81560548>] (u64_stats_fetch_begin include/linux/u64_stats_sync.h:175 [inline])
 [<802beca4>] (lock_acquire) from [<81560548>] (nsim_get_stats64+0xdc/0xf0 drivers/net/netdevsim/netdev.c:70)
 [<8156046c>] (nsim_get_stats64) from [<81e2efa0>] (dev_get_stats+0x44/0xd0 net/core/dev.c:10405)
 [<81e2ef5c>] (dev_get_stats) from [<81e53204>] (rtnl_fill_stats+0x38/0x120 net/core/rtnetlink.c:1211)
 [<81e531cc>] (rtnl_fill_stats) from [<81e59d58>] (rtnl_fill_ifinfo+0x6d4/0x148c net/core/rtnetlink.c:1783)
 [<81e59684>] (rtnl_fill_ifinfo) from [<81e5ceb4>] (rtmsg_ifinfo_build_skb+0x9c/0x108 net/core/rtnetlink.c:3798)
 [<81e5ce18>] (rtmsg_ifinfo_build_skb) from [<81e5d0ac>] (rtmsg_ifinfo_event net/core/rtnetlink.c:3830 [inline])
 [<81e5ce18>] (rtmsg_ifinfo_build_skb) from [<81e5d0ac>] (rtmsg_ifinfo_event net/core/rtnetlink.c:3821 [inline])
 [<81e5ce18>] (rtmsg_ifinfo_build_skb) from [<81e5d0ac>] (rtmsg_ifinfo+0x44/0x70 net/core/rtnetlink.c:3839)
 [<81e5d068>] (rtmsg_ifinfo) from [<81e45c2c>] (register_netdevice+0x664/0x68c net/core/dev.c:10103)
 [<81e455c8>] (register_netdevice) from [<815608bc>] (nsim_create+0xf8/0x124 drivers/net/netdevsim/netdev.c:317)
 [<815607c4>] (nsim_create) from [<81561184>] (__nsim_dev_port_add+0x108/0x188 drivers/net/netdevsim/dev.c:941)
 [<8156107c>] (__nsim_dev_port_add) from [<815620d8>] (nsim_dev_port_add_all drivers/net/netdevsim/dev.c:990 [inline])
 [<8156107c>] (__nsim_dev_port_add) from [<815620d8>] (nsim_dev_probe+0x5cc/0x750 drivers/net/netdevsim/dev.c:1119)
 [<81561b0c>] (nsim_dev_probe) from [<815661dc>] (nsim_bus_probe+0x10/0x14 drivers/net/netdevsim/bus.c:287)
 [<815661cc>] (nsim_bus_probe) from [<811724c0>] (really_probe+0x100/0x50c drivers/base/dd.c:554)
 [<811723c0>] (really_probe) from [<811729c4>] (driver_probe_device+0xf8/0x1c8 drivers/base/dd.c:740)
 [<811728cc>] (driver_probe_device) from [<81172fe4>] (__device_attach_driver+0x8c/0xf0 drivers/base/dd.c:846)
 [<81172f58>] (__device_attach_driver) from [<8116fee0>] (bus_for_each_drv+0x88/0xd8 drivers/base/bus.c:431)
 [<8116fe58>] (bus_for_each_drv) from [<81172c6c>] (__device_attach+0xdc/0x1d0 drivers/base/dd.c:914)
 [<81172b90>] (__device_attach) from [<8117305c>] (device_initial_probe+0x14/0x18 drivers/base/dd.c:961)
 [<81173048>] (device_initial_probe) from [<81171358>] (bus_probe_device+0x90/0x98 drivers/base/bus.c:491)
 [<811712c8>] (bus_probe_device) from [<8116e77c>] (device_add+0x320/0x824 drivers/base/core.c:3109)
 [<8116e45c>] (device_add) from [<8116ec9c>] (device_register+0x1c/0x20 drivers/base/core.c:3182)
 [<8116ec80>] (device_register) from [<81566710>] (nsim_bus_dev_new drivers/net/netdevsim/bus.c:336 [inline])
 [<8116ec80>] (device_register) from [<81566710>] (new_device_store+0x178/0x208 drivers/net/netdevsim/bus.c:215)
 [<81566598>] (new_device_store) from [<8116fcb4>] (bus_attr_store+0x2c/0x38 drivers/base/bus.c:122)
 [<8116fc88>] (bus_attr_store) from [<805b4b8c>] (sysfs_kf_write+0x48/0x54 fs/sysfs/file.c:139)
 [<805b4b44>] (sysfs_kf_write) from [<805b3c90>] (kernfs_fop_write_iter+0x128/0x1ec fs/kernfs/file.c:296)
 [<805b3b68>] (kernfs_fop_write_iter) from [<804d22fc>] (call_write_iter include/linux/fs.h:1901 [inline])
 [<805b3b68>] (kernfs_fop_write_iter) from [<804d22fc>] (new_sync_write fs/read_write.c:518 [inline])
 [<805b3b68>] (kernfs_fop_write_iter) from [<804d22fc>] (vfs_write+0x3dc/0x57c fs/read_write.c:605)
 [<804d1f20>] (vfs_write) from [<804d2604>] (ksys_write+0x68/0xec fs/read_write.c:658)
 [<804d259c>] (ksys_write) from [<804d2698>] (__do_sys_write fs/read_write.c:670 [inline])
 [<804d259c>] (ksys_write) from [<804d2698>] (sys_write+0x10/0x14 fs/read_write.c:667)
 [<804d2688>] (sys_write) from [<80200060>] (ret_fast_syscall+0x0/0x2c arch/arm/mm/proc-v7.S:64)

Fixes: 83c9e13aa39a ("netdevsim: add software driver for testing offloads")
Reported-by: syzbot+e74a6857f2d0efe3ad81@syzkaller.appspotmail.com
Tested-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Hillf Danton <hdanton@sina.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: usb: qmi_wwan: allow qmimux add/del with master up

commit 6c59cff38e66584ae3ac6c2f0cbd8d039c710ba7 upstream.

There's no reason for preventing the creation and removal
of qmimux network interfaces when the underlying interface
is up.

This makes qmi_wwan mux implementation more similar to the
rmnet one, simplifying userspace management of the same
logical interfaces.

Fixes: c6adf77953bc ("net: usb: qmi_wwan: add qmap mux protocol support")
Reported-by: Aleksander Morgado <aleksander@aleksander.es>
Signed-off-by: Daniele Palmas <dnlplm@gmail.com>
Acked-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: sched: avoid duplicates in classes dump

commit bfc2560563586372212b0a8aeca7428975fa91fe upstream.

This is a follow up of commit ea3274695353 ("net: sched: avoid
duplicates in qdisc dump") which has fixed the issue only for the qdisc
dump.

The duplicate printing also occurs when dumping the classes via
  tc class show dev eth0

Fixes: 59cc1f61f09c ("net: sched: convert qdisc linked list to hashtable")
Signed-off-by: Maximilian Heyne <mheyne@amazon.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nexthop: Do not flush blackhole nexthops when loopback goes down

commit 76c03bf8e2624076b88d93542d78e22d5345c88e upstream.

As far as user space is concerned, blackhole nexthops do not have a
nexthop device and therefore should not be affected by the
administrative or carrier state of any netdev.

However, when the loopback netdev goes down all the blackhole nexthops
are flushed. This happens because internally the kernel associates
blackhole nexthops with the loopback netdev.

This behavior is both confusing to those not familiar with kernel
internals and also diverges from the legacy API where blackhole IPv4
routes are not flushed when the loopback netdev goes down:

 # ip route add blackhole 198.51.100.0/24
 # ip link set dev lo down
 # ip route show 198.51.100.0/24
 blackhole 198.51.100.0/24

Blackhole IPv6 routes are flushed, but at least user space knows that
they are associated with the loopback netdev:

 # ip -6 route show 2001:db8:1::/64
 blackhole 2001:db8:1::/64 dev lo metric 1024 pref medium

Fix this by only flushing blackhole nexthops when the loopback netdev is
unregistered.

Fixes: ab84be7e54fc ("net: Initial nexthop code")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reported-by: Donald Sharp <sharpd@nvidia.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: stmmac: fix incorrect DMA channel intr enable setting of EQoS v4.10

commit 879c348c35bb5fb758dd881d8a97409c1862dae8 upstream.

We introduce dwmac410_dma_init_channel() here for both EQoS v4.10 and
above which use different DMA_CH(n)_Interrupt_Enable bit definitions for
NIE and AIE.

Fixes: 48863ce5940f ("stmmac: add DMA support for GMAC 4.xx")
Signed-off-by: Ong Boon Leong <boon.leong.ong@intel.com>
Signed-off-by: Ramesh Babu B <ramesh.babu.b@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/mlx4_en: update moderation when config reset

commit 00ff801bb8ce6711e919af4530b6ffa14a22390a upstream.

This patch fixes a bug that the moderation config will not be
applied when calling mlx4_en_reset_config. For example, when
turning on rx timestamping, mlx4_en_reset_config() will be called,
causing the NIC to forget previous moderation config.

This fix is in phase with a previous fix:
commit 79c54b6bbf06 ("net/mlx4_en: Fix TX moderation info loss
after set_ringparam is called")

Tested: Before this patch, on a host with NIC using mlx4, run
netserver and stream TCP to the host at full utilization.
$ sar -I SUM 1
                 INTR    intr/s
14:03:56          sum  48758.00

After rx hwtstamp is enabled:
$ sar -I SUM 1
14:10:38          sum 317771.00
We see the moderation is not working properly and issued 7x more
interrupts.

After the patch, and turned on rx hwtstamp, the rate of interrupts
is as expected:
$ sar -I SUM 1
14:52:11          sum  49332.00

Fixes: 79c54b6bbf06 ("net/mlx4_en: Fix TX moderation info loss after set_ringparam is called")
Signed-off-by: Kevin(Yudong) Yang <yyd@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
CC: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: enetc: don't overwrite the RSS indirection table when initializing

commit c646d10dda2dcde82c6ce5a474522621ab2b8b19 upstream.

After the blamed patch, all RX traffic gets hashed to CPU 0 because the
hashing indirection table set up in:

enetc_pf_probe
-> enetc_alloc_si_resources
   -> enetc_configure_si
      -> enetc_setup_default_rss_table

is overwritten later in:

enetc_pf_probe
-> enetc_init_port_rss_memory

which zero-initializes the entire port RSS table in order to avoid ECC errors.

The trouble really is that enetc_init_port_rss_memory really neads
enetc_alloc_si_resources to be called, because it depends upon
enetc_alloc_cbdr and enetc_setup_cbdr. But that whole enetc_configure_si
thing could have been better thought out, it has nothing to do in a
function called "alloc_si_resources", especially since its counterpart,
"free_si_resources", does nothing to unwind the configuration of the SI.

The point is, we need to pull out enetc_configure_si out of
enetc_alloc_resources, and move it after enetc_init_port_rss_memory.
This allows us to set up the default RSS indirection table after
initializing the memory.

Fixes: 07bf34a50e32 ("net: enetc: initialize the RFS and RSS memories")
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "mm, slub: consider rest of partial list if acquire_slab() fails"

commit 9b1ea29bc0d7b94d420f96a0f4121403efc3dd85 upstream.

This reverts commit 8ff60eb052eeba95cfb3efe16b08c9199f8121cf.

The kernel test robot reports a huge performance regression due to the
commit, and the reason seems fairly straightforward: when there is
contention on the page list (which is what causes acquire_slab() to
fail), we do _not_ want to just loop and try again, because that will
transfer the contention to the 'n->list_lock' spinlock we hold, and
just make things even worse.

This is admittedly likely a problem only on big machines - the kernel
test robot report comes from a 96-thread dual socket Intel Xeon Gold
6252 setup, but the regression there really is quite noticeable:

   -47.9% regression of stress-ng.rawpkt.ops_per_sec

and the commit that was marked as being fixed (7ced37197196: "slub:
Acquire_slab() avoid loop") actually did the loop exit early very
intentionally (the hint being that "avoid loop" part of that commit
message), exactly to avoid this issue.

The correct thing to do may be to pick some kind of reasonable middle
ground: instead of breaking out of the loop on the very first sign of
contention, or trying over and over and over again, the right thing may
be to re-try _once_, and then give up on the second failure (or pick
your favorite value for "once"..).

Reported-by: kernel test robot <oliver.sang@intel.com>
Link: https://lore.kernel.org/lkml/20210301080404.GF12822@xsang-OptiPlex-9020/
Cc: Jann Horn <jannh@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cifs: return proper error code in statfs(2)

commit 14302ee3301b3a77b331cc14efb95bf7184c73cc upstream.

In cifs_statfs(), if server->ops->queryfs is not NULL, then we should
use its return value rather than always returning 0.  Instead, use rc
variable as it is properly set to 0 in case there is no
server->ops->queryfs.

Signed-off-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
CC: <stable@vger.kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mount: fix mounting of detached mounts onto targets that reside on shared mounts

commit ee2e3f50629f17b0752b55b2566c15ce8dafb557 upstream.

Creating a series of detached mounts, attaching them to the filesystem,
and unmounting them can be used to trigger an integer overflow in
ns->mounts causing the kernel to block any new mounts in count_mounts()
and returning ENOSPC because it falsely assumes that the maximum number
of mounts in the mount namespace has been reached, i.e. it thinks it
can't fit the new mounts into the mount namespace anymore.

Depending on the number of mounts in your system, this can be reproduced
on any kernel that supportes open_tree() and move_mount() by compiling
and running the following program:

  /* SPDX-License-Identifier: LGPL-2.1+ */

  #define _GNU_SOURCE
  #include <errno.h>
  #include <fcntl.h>
  #include <getopt.h>
  #include <limits.h>
  #include <stdbool.h>
  #include <stdio.h>
  #include <stdlib.h>
  #include <string.h>
  #include <sys/mount.h>
  #include <sys/stat.h>
  #include <sys/syscall.h>
  #include <sys/types.h>
  #include <unistd.h>

  /* open_tree() */
  #ifndef OPEN_TREE_CLONE
  #define OPEN_TREE_CLONE 1
  #endif

  #ifndef OPEN_TREE_CLOEXEC
  #define OPEN_TREE_CLOEXEC O_CLOEXEC
  #endif

  #ifndef __NR_open_tree
          #if defined __alpha__
                  #define __NR_open_tree 538
          #elif defined _MIPS_SIM
                  #if _MIPS_SIM == _MIPS_SIM_ABI32        /* o32 */
                          #define __NR_open_tree 4428
                  #endif
                  #if _MIPS_SIM == _MIPS_SIM_NABI32       /* n32 */
                          #define __NR_open_tree 6428
                  #endif
                  #if _MIPS_SIM == _MIPS_SIM_ABI64        /* n64 */
                          #define __NR_open_tree 5428
                  #endif
          #elif defined __ia64__
                  #define __NR_open_tree (428 + 1024)
          #else
                  #define __NR_open_tree 428
          #endif
  #endif

  /* move_mount() */
  #ifndef MOVE_MOUNT_F_EMPTY_PATH
  #define MOVE_MOUNT_F_EMPTY_PATH 0x00000004 /* Empty from path permitted */
  #endif

  #ifndef __NR_move_mount
          #if defined __alpha__
                  #define __NR_move_mount 539
          #elif defined _MIPS_SIM
                  #if _MIPS_SIM == _MIPS_SIM_ABI32        /* o32 */
                          #define __NR_move_mount 4429
                  #endif
                  #if _MIPS_SIM == _MIPS_SIM_NABI32       /* n32 */
                          #define __NR_move_mount 6429
                  #endif
                  #if _MIPS_SIM == _MIPS_SIM_ABI64        /* n64 */
                          #define __NR_move_mount 5429
                  #endif
          #elif defined __ia64__
                  #define __NR_move_mount (428 + 1024)
          #else
                  #define __NR_move_mount 429
          #endif
  #endif

  static inline int sys_open_tree(int dfd, const char *filename, unsigned int flags)
  {
          return syscall(__NR_open_tree, dfd, filename, flags);
  }

  static inline int sys_move_mount(int from_dfd, const char *from_pathname, int to_dfd,
                                   const char *to_pathname, unsigned int flags)
  {
          return syscall(__NR_move_mount, from_dfd, from_pathname, to_dfd, to_pathname, flags);
  }

  static bool is_shared_mountpoint(const char *path)
  {
          bool shared = false;
          FILE *f = NULL;
          char *line = NULL;
          int i;
          size_t len = 0;

          f = fopen("/proc/self/mountinfo", "re");
          if (!f)
                  return 0;

          while (getline(&line, &len, f) > 0) {
                  char *slider1, *slider2;

                  for (slider1 = line, i = 0; slider1 && i < 4; i++)
                          slider1 = strchr(slider1 + 1, ' ');

                  if (!slider1)
                          continue;

                  slider2 = strchr(slider1 + 1, ' ');
                  if (!slider2)
                          continue;

                  *slider2 = '\0';
                  if (strcmp(slider1 + 1, path) == 0) {
                          /* This is the path. Is it shared? */
                          slider1 = strchr(slider2 + 1, ' ');
                          if (slider1 && strstr(slider1, "shared:")) {
                                  shared = true;
                                  break;
                          }
                  }
          }
          fclose(f);
          free(line);

          return shared;
  }

  static void usage(void)
  {
          const char *text = "mount-new [--recursive] <base-dir>\n";
          fprintf(stderr, "%s", text);
          _exit(EXIT_SUCCESS);
  }

  #define exit_usage(format, ...)                              \
          ({                                                   \
                  fprintf(stderr, format "\n", ##__VA_ARGS__); \
                  usage();                                     \
          })

  #define exit_log(format, ...)                                \
          ({                                                   \
                  fprintf(stderr, format "\n", ##__VA_ARGS__); \
                  exit(EXIT_FAILURE);                          \
          })

  static const struct option longopts[] = {
          {"help",        no_argument,            0,      'a'},
          { NULL,         no_argument,            0,       0 },
  };

  int main(int argc, char *argv[])
  {
          int exit_code = EXIT_SUCCESS, index = 0;
          int dfd, fd_tree, new_argc, ret;
          char *base_dir;
          char *const *new_argv;
          char target[PATH_MAX];

          while ((ret = getopt_long_only(argc, argv, "", longopts, &index)) != -1) {
                  switch (ret) {
                  case 'a':
                          /* fallthrough */
                  default:
                          usage();
                  }
          }

          new_argv = &argv[optind];
          new_argc = argc - optind;
          if (new_argc < 1)
                  exit_usage("Missing base directory\n");
          base_dir = new_argv[0];

          if (*base_dir != '/')
                  exit_log("Please specify an absolute path");

          /* Ensure that target is a shared mountpoint. */
          if (!is_shared_mountpoint(base_dir))
                  exit_log("Please ensure that \"%s\" is a shared mountpoint", base_dir);

          dfd = open(base_dir, O_RDONLY | O_DIRECTORY | O_CLOEXEC);
          if (dfd < 0)
                  exit_log("%m - Failed to open base directory \"%s\"", base_dir);

          ret = mkdirat(dfd, "detached-move-mount", 0755);
          if (ret < 0)
                  exit_log("%m - Failed to create required temporary directories");

          ret = snprintf(target, sizeof(target), "%s/detached-move-mount", base_dir);
          if (ret < 0 || (size_t)ret >= sizeof(target))
                  exit_log("%m - Failed to assemble target path");

          /*
           * Having a mount table with 10000 mounts is already quite excessive
           * and shoult account even for weird test systems.
           */
          for (size_t i = 0; i < 10000; i++) {
                  fd_tree = sys_open_tree(dfd, "detached-move-mount",
                                          OPEN_TREE_CLONE |
                                          OPEN_TREE_CLOEXEC |
                                          AT_EMPTY_PATH);
                  if (fd_tree < 0) {
                          fprintf(stderr, "%m - Failed to open %d(detached-move-mount)", dfd);
                          exit_code = EXIT_FAILURE;
                          break;
                  }

                  ret = sys_move_mount(fd_tree, "", dfd, "detached-move-mount", MOVE_MOUNT_F_EMPTY_PATH);
                  if (ret < 0) {
                          if (errno == ENOSPC)
                                  fprintf(stderr, "%m - Buggy mount counting");
                          else
                                  fprintf(stderr, "%m - Failed to attach mount to %d(detached-move-mount)", dfd);
                          exit_code = EXIT_FAILURE;
                          break;
                  }
                  close(fd_tree);

                  ret = umount2(target, MNT_DETACH);
                  if (ret < 0) {
                          fprintf(stderr, "%m - Failed to unmount %s", target);
                          exit_code = EXIT_FAILURE;
                          break;
                  }
          }

          (void)unlinkat(dfd, "detached-move-mount", AT_REMOVEDIR);
          close(dfd);

          exit(exit_code);
  }

and wait for the kernel to refuse any new mounts by returning ENOSPC.
How many iterations are needed depends on the number of mounts in your
system. Assuming you have something like 50 mounts on a standard system
it should be almost instantaneous.

The root cause of this is that detached mounts aren't handled correctly
when source and target mount are identical and reside on a shared mount
causing a broken mount tree where the detached source itself is
propagated which propagation prevents for regular bind-mounts and new
mounts. This ultimately leads to a miscalculation of the number of
mounts in the mount namespace.

Detached mounts created via
open_tree(fd, path, OPEN_TREE_CLONE)
are essentially like an unattached new mount, or an unattached
bind-mount. They can then later on be attached to the filesystem via
move_mount() which calls into attach_recursive_mount(). Part of
attaching it to the filesystem is making sure that mounts get correctly
propagated in case the destination mountpoint is MS_SHARED, i.e. is a
shared mountpoint. This is done by calling into propagate_mnt() which
walks the list of peers calling propagate_one() on each mount in this
list making sure it receives the propagation event.
The propagate_one() functions thereby skips both new mounts and bind
mounts to not propagate them "into themselves". Both are identified by
checking whether the mount is already attached to any mount namespace in
mnt->mnt_ns. The is what the IS_MNT_NEW() helper is responsible for.

However, detached mounts have an anonymous mount namespace attached to
them stashed in mnt->mnt_ns which means that IS_MNT_NEW() doesn't
realize they need to be skipped causing the mount to propagate "into
itself" breaking the mount table and causing a disconnect between the
number of mounts recorded as being beneath or reachable from the target
mountpoint and the number of mounts actually recorded/counted in
ns->mounts ultimately causing an overflow which in turn prevents any new
mounts via the ENOSPC issue.

So teach propagation to handle detached mounts by making it aware of
them. I've been tracking this issue down for the last couple of days and
then verifying that the fix is correct by
unmounting everything in my current mount table leaving only /proc and
/sys mounted and running the reproducer above overnight verifying the
number of mounts counted in ns->mounts. With this fix the counts are
correct and the ENOSPC issue can't be reproduced.

This change will only have an effect on mounts created with the new
mount API since detached mounts cannot be created with the old mount API
so regressions are extremely unlikely.

Link: https://lore.kernel.org/r/20210306101010.243666-1-christian.brauner@ubuntu.com
Fixes: 2db154b3ea8e ("vfs: syscall: Add move_mount(2) to move mounts around")
Cc: David Howells <dhowells@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: linux-fsdevel@vger.kernel.org
Cc: <stable@vger.kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

powerpc/603: Fix protection of user pages mapped with PROT_NONE

commit c119565a15a628efdfa51352f9f6c5186e506a1c upstream.

On book3s/32, page protection is defined by the PP bits in the PTE
which provide the following protection depending on the access
keys defined in the matching segment register:
- PP 00 means RW with key 0 and N/A with key 1.
- PP 01 means RW with key 0 and RO with key 1.
- PP 10 means RW with both key 0 and key 1.
- PP 11 means RO with both key 0 and key 1.

Since the implementation of kernel userspace access protection,
PP bits have been set as follows:
- PP00 for pages without _PAGE_USER
- PP01 for pages with _PAGE_USER and _PAGE_RW
- PP11 for pages with _PAGE_USER and without _PAGE_RW

For kernelspace segments, kernel accesses are performed with key 0
and user accesses are performed with key 1. As PP00 is used for
non _PAGE_USER pages, user can't access kernel pages not flagged
_PAGE_USER while kernel can.

For userspace segments, both kernel and user accesses are performed
with key 0, therefore pages not flagged _PAGE_USER are still
accessible to the user.

This shouldn't be an issue, because userspace is expected to be
accessible to the user. But unlike most other architectures, powerpc
implements PROT_NONE protection by removing _PAGE_USER flag instead of
flagging the page as not valid. This means that pages in userspace
that are not flagged _PAGE_USER shall remain inaccessible.

To get the expected behaviour, just mimic other architectures in the
TLB miss handler by checking _PAGE_USER permission on userspace
accesses as if it was the _PAGE_PRESENT bit.

Note that this problem only is only for 603 cores. The 604+ have
an hash table, and hash_page() function already implement the
verification of _PAGE_USER permission on userspace pages.

Fixes: f342adca3afc ("powerpc/32s: Prepare Kernel Userspace Access Protection")
Cc: stable@vger.kernel.org # v5.2+
Reported-by: Christoph Plattner <christoph.plattner@thalesgroup.com>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/4a0c6e3bb8f0c162457bf54d9bc6fd8d7b55129f.1612160907.git.christophe.leroy@csgroup.eu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mt76: dma: do not report truncated frames to mac80211

commit d0bd52c591a1070c54dc428e926660eb4f981099 upstream.

Commit b102f0c522cf6 ("mt76: fix array overflow on receiving too many
fragments for a packet") fixes a possible OOB access but it introduces a
memory leak since the pending frame is not released to page_frag_cache
if the frag array of skb_shared_info is full. Commit 93a1d4791c10
("mt76: dma: fix a possible memory leak in mt76_add_fragment()") fixes
the issue but does not free the truncated skb that is forwarded to
mac80211 layer. Fix the leftover issue discarding even truncated skbs.

Fixes: 93a1d4791c10 ("mt76: dma: fix a possible memory leak in mt76_add_fragment()")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/a03166fcc8214644333c68674a781836e0f57576.1612697217.git.lorenzo@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ibmvnic: always store valid MAC address

commit 67eb211487f0c993d9f402d1c196ef159fd6a3b5 upstream.

The last change to ibmvnic_set_mac(), 8fc3672a8ad3, meant to prevent
users from setting an invalid MAC address on an ibmvnic interface
that has not been brought up yet. The change also prevented the
requested MAC address from being stored by the adapter object for an
ibmvnic interface when the state of the ibmvnic interface is
VNIC_PROBED - that is after probing has finished but before the
ibmvnic interface is brought up. The MAC address stored by the
adapter object is used and sent to the hypervisor for checking when
an ibmvnic interface is brought up.

The ibmvnic driver ignoring the requested MAC address when in
VNIC_PROBED state caused LACP bonds (bonds in 802.3ad mode) with more
than one slave to malfunction. The bonding code must be able to
change the MAC address of its slaves before they are brought up
during enslaving. The inability of kernels with 8fc3672a8ad3 to set
the MAC addresses of bonding slaves is observable in the output of
"ip address show". The MAC addresses of the slaves are the same as
the MAC address of the bond on a working system whereas the slaves
retain their original MAC addresses on a system with a malfunctioning
LACP bond.

Fixes: 8fc3672a8ad3 ("ibmvnic: fix ibmvnic_set_mac")
Signed-off-by: Jiri Wiesner <jwiesner@suse.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

samples, bpf: Add missing munmap in xdpsock

commit 6bc6699881012b5bd5d49fa861a69a37fc01b49c upstream.

We mmap the umem region, but we never munmap it.
Add the missing call at the end of the cleanup.

Fixes: 3945b37a975d ("samples/bpf: use hugepages in xdpsock app")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/20210303185636.18070-3-maciej.fijalkowski@intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests/bpf: Mask bpf_csum_diff() return value to 16 bits in test_verifier

commit 6185266c5a853bb0f2a459e3ff594546f277609b upstream.

The verifier test labelled "valid read map access into a read-only array
2" calls the bpf_csum_diff() helper and checks its return value. However,
architecture implementations of csum_partial() (which is what the helper
uses) differ in whether they fold the return value to 16 bit or not. For
example, x86 version has ...

if (unlikely(odd)) {
result = from32to16(result);
result = ((result >> 8) & 0xff) | ((result & 0xff) << 8);
}

... while generic lib/checksum.c does:

result = from32to16(result);
if (odd)
result = ((result >> 8) & 0xff) | ((result & 0xff) << 8);

This makes the helper return different values on different architectures,
breaking the test on non-x86. To fix this, add an additional instruction
to always mask the return value to 16 bits, and update the expected return
value accordingly.

Fixes: fb2abb73e575 ("bpf, selftest: test {rd, wr}only flags and direct value access")
Signed-off-by: Yauheni Kaliuta <yauheni.kaliuta@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20210228103017.320240-1-yauheni.kaliuta@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests/bpf: No need to drop the packet when there is no geneve opt

commit 557c223b643a35effec9654958d8edc62fd2603a upstream.

In bpf geneve tunnel test we set geneve option on tx side. On rx side we
only call bpf_skb_get_tunnel_opt(). Since commit 9c2e14b48119 ("ip_tunnels:
Set tunnel option flag when tunnel metadata is present") geneve_rx() will
not add TUNNEL_GENEVE_OPT flag if there is no geneve option, which cause
bpf_skb_get_tunnel_opt() return ENOENT and _geneve_get_tunnel() in
test_tunnel_kern.c drop the packet.

As it should be valid that bpf_skb_get_tunnel_opt() return error when
there is not tunnel option, there is no need to drop the packet and
break all geneve rx traffic. Just set opt_class to 0 in this test and
keep returning TC_ACT_OK.

Fixes: 933a741e3b82 ("selftests/bpf: bpf tunnel test.")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: William Tu <u9012063@gmail.com>
Link: https://lore.kernel.org/bpf/20210224081403.1425474-1-liuhangbin@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

netfilter: x_tables: gpf inside xt_find_revision()

commit 8e24edddad152b998b37a7f583175137ed2e04a5 upstream.

nested target/match_revfn() calls work with xt[NFPROTO_UNSPEC] lists
without taking xt[NFPROTO_UNSPEC].mutex. This can race with module unload
and cause host to crash:

general protection fault: 0000 [#1]
Modules linked in: ... [last unloaded: xt_cluster]
CPU: 0 PID: 542455 Comm: iptables
RIP: 0010:[<ffffffff8ffbd518>]  [<ffffffff8ffbd518>] strcmp+0x18/0x40
RDX: 0000000000000003 RSI: ffff9a5a5d9abe10 RDI: dead000000000111
R13: ffff9a5a5d9abe10 R14: ffff9a5a5d9abd8c R15: dead000000000100
(VvS: %R15 -- &xt_match,  %RDI -- &xt_match.name,
xt_cluster unregister match in xt[NFPROTO_UNSPEC].match list)
Call Trace:
 [<ffffffff902ccf44>] match_revfn+0x54/0xc0
 [<ffffffff902ccf9f>] match_revfn+0xaf/0xc0
 [<ffffffff902cd01e>] xt_find_revision+0x6e/0xf0
 [<ffffffffc05a5be0>] do_ipt_get_ctl+0x100/0x420 [ip_tables]
 [<ffffffff902cc6bf>] nf_getsockopt+0x4f/0x70
 [<ffffffff902dd99e>] ip_getsockopt+0xde/0x100
 [<ffffffff903039b5>] raw_getsockopt+0x25/0x50
 [<ffffffff9026c5da>] sock_common_getsockopt+0x1a/0x20
 [<ffffffff9026b89d>] SyS_getsockopt+0x7d/0xf0
 [<ffffffff903cbf92>] system_call_fastpath+0x25/0x2a

Fixes: 656caff20e1 ("netfilter 04/09: x_tables: fix match/target revision lookup")
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Reviewed-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

netfilter: nf_nat: undo erroneous tcp edemux lookup

commit 03a3ca37e4c6478e3a84f04c8429dd5889e107fd upstream.

Under extremely rare conditions TCP early demux will retrieve the wrong
socket.

1. local machine establishes a connection to a remote server, S, on port
   p.

   This gives:
   laddr:lport -> S:p
   ... both in tcp and conntrack.

2. local machine establishes a connection to host H, on port p2.
   2a. TCP stack choses same laddr:lport, so we have
   laddr:lport -> H:p2 from TCP point of view.
   2b). There is a destination NAT rewrite in place, translating
        H:p2 to S:p.  This results in following conntrack entries:

   I)  laddr:lport -> S:p  (origin)  S:p -> laddr:lport (reply)
   II) laddr:lport -> H:p2 (origin)  S:p -> laddr:lport2 (reply)

   NAT engine has rewritten laddr:lport to laddr:lport2 to map
   the reply packet to the correct origin.

   When server sends SYN/ACK to laddr:lport2, the PREROUTING hook
   will undo-the SNAT transformation, rewriting IP header to
   S:p -> laddr:lport

   This causes TCP early demux to associate the skb with the TCP socket
   of the first connection.

   The INPUT hook will then reverse the DNAT transformation, rewriting
   the IP header to H:p2 -> laddr:lport.

Because packet ends up with the wrong socket, the new connection
never completes: originator stays in SYN_SENT and conntrack entry
remains in SYN_RECV until timeout, and responder retransmits SYN/ACK
until it gives up.

To resolve this, orphan the skb after the input rewrite:
Because the source IP address changed, the socket must be incorrect.
We can't move the DNAT undo to prerouting due to backwards
compatibility, doing so will make iptables/nftables rules to no longer
match the way they did.

After orphan, the packet will be handed to the next protocol layer
(tcp, udp, ...) and that will repeat the socket lookup just like as if
early demux was disabled.

Fixes: 41063e9dd1195 ("ipv4: Early TCP socket demux.")
Closes: https://bugzilla.netfilter.org/show_bug.cgi?id=1427
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>