linux-2.6-microblaze.git
4 years agousb: dwc2: Fix error path in gadget registration
Marek Szyprowski [Thu, 16 Jul 2020 12:09:48 +0000 (14:09 +0200)]
usb: dwc2: Fix error path in gadget registration

When gadget registration fails, one should not call usb_del_gadget_udc().
Ensure this by setting gadget->udc to NULL. Also in case of a failure
there is no need to disable low-level hardware, so return immiedetly
instead of jumping to error_init label.

This fixes the following kernel NULL ptr dereference on gadget failure
(can be easily triggered with g_mass_storage without any module
parameters):

dwc2 12480000.hsotg: dwc2_check_params: Invalid parameter besl=1
dwc2 12480000.hsotg: dwc2_check_params: Invalid parameter g_np_tx_fifo_size=1024
dwc2 12480000.hsotg: EPs: 16, dedicated fifos, 7808 entries in SPRAM
Mass Storage Function, version: 2009/09/11
LUN: removable file: (no medium)
no file given for LUN0
g_mass_storage 12480000.hsotg: failed to start g_mass_storage: -22
8<--- cut here ---
Unable to handle kernel NULL pointer dereference at virtual address 00000104
pgd = (ptrval)
[00000104] *pgd=00000000
Internal error: Oops: 805 [#1] PREEMPT SMP ARM
Modules linked in:
CPU: 0 PID: 12 Comm: kworker/0:1 Not tainted 5.8.0-rc5 #3133
Hardware name: Samsung Exynos (Flattened Device Tree)
Workqueue: events deferred_probe_work_func
PC is at usb_del_gadget_udc+0x38/0xc4
LR is at __mutex_lock+0x31c/0xb18
...
Process kworker/0:1 (pid: 12, stack limit = 0x(ptrval))
Stack: (0xef121db0 to 0xef122000)
...
[<c076bf3c>] (usb_del_gadget_udc) from [<c0726bec>] (dwc2_hsotg_remove+0x10/0x20)
[<c0726bec>] (dwc2_hsotg_remove) from [<c0711208>] (dwc2_driver_probe+0x57c/0x69c)
[<c0711208>] (dwc2_driver_probe) from [<c06247c0>] (platform_drv_probe+0x6c/0xa4)
[<c06247c0>] (platform_drv_probe) from [<c0621df4>] (really_probe+0x200/0x48c)
[<c0621df4>] (really_probe) from [<c06221e8>] (driver_probe_device+0x78/0x1fc)
[<c06221e8>] (driver_probe_device) from [<c061fcd4>] (bus_for_each_drv+0x74/0xb8)
[<c061fcd4>] (bus_for_each_drv) from [<c0621b54>] (__device_attach+0xd4/0x16c)
[<c0621b54>] (__device_attach) from [<c0620c98>] (bus_probe_device+0x88/0x90)
[<c0620c98>] (bus_probe_device) from [<c06211b0>] (deferred_probe_work_func+0x3c/0xd0)
[<c06211b0>] (deferred_probe_work_func) from [<c0149280>] (process_one_work+0x234/0x7dc)
[<c0149280>] (process_one_work) from [<c014986c>] (worker_thread+0x44/0x51c)
[<c014986c>] (worker_thread) from [<c0150b1c>] (kthread+0x158/0x1a0)
[<c0150b1c>] (kthread) from [<c0100114>] (ret_from_fork+0x14/0x20)
Exception stack(0xef121fb0 to 0xef121ff8)
...
---[ end trace 9724c2fc7cc9c982 ]---

While fixing this also fix the double call to dwc2_lowlevel_hw_disable()
if dr_mode is set to USB_DR_MODE_PERIPHERAL. In such case low-level
hardware is already disabled before calling usb_add_gadget_udc(). That
function correctly preserves low-level hardware state, there is no need
for the second unconditional dwc2_lowlevel_hw_disable() call.

Fixes: 207324a321a8 ("usb: dwc2: Postponed gadget registration to the udc class driver")
Acked-by: Minas Harutyunyan <hminas@synopsys.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
4 years agousb: gadget: udc: atmel: implement .pullup callback
Michał Mirosław [Wed, 8 Jul 2020 18:04:10 +0000 (20:04 +0200)]
usb: gadget: udc: atmel: implement .pullup callback

Implement udc->pullup callback, so that udc_connect/disconnect work.
This is needed for composite gadget, as it assumes udc_disconnect()
actually works and calls its ->disconnect callback.

Acked-by: Cristian Birsan <cristian.birsan@microchip.com>
Acked-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
4 years agousb: gadget: udc: Flush pending work also in error path
Marek Szyprowski [Thu, 16 Jul 2020 11:55:25 +0000 (13:55 +0200)]
usb: gadget: udc: Flush pending work also in error path

When binding an UDC driver to the pending gadget fails in
check_pending_gadget_drivers(), the usb_add_gadget_udc_release() function
ends without waiting for the usb_gadget_state_work to finish, what in
turn might cause the whole struct usb_gadget being freed by the caller
before the usb_gadget_state_work being executed.

This can be observed on some boards with USB Mass Storage gadget
compiled-in and kernel booted without the needed module parameters:

dwc2 12480000.hsotg: dwc2_check_params: Invalid parameter besl=1
dwc2 12480000.hsotg: dwc2_check_params: Invalid parameter g_np_tx_fifo_size=1024
dwc2 12480000.hsotg: EPs: 16, dedicated fifos, 7808 entries in SPRAM
Mass Storage Function, version: 2009/09/11
LUN: removable file: (no medium)
no file given for LUN0
g_mass_storage 12480000.hsotg: failed to start g_mass_storage: -22
dwc2: probe of 12480000.hsotg failed with error -22
8<--- cut here ---
Unable to handle kernel NULL pointer dereference at virtual address 00000004
pgd = (ptrval)
[00000004] *pgd=00000000
Internal error: Oops: 5 [#1] PREEMPT SMP ARM
Modules linked in:
CPU: 1 PID: 88 Comm: kworker/1:2 Not tainted 5.8.0-rc5-next-20200715-00062-gc5bb489ae825-dirty #8792
Hardware name: Samsung Exynos (Flattened Device Tree)
Workqueue:  0x0 (rcu_gp)
PC is at process_one_work+0x44/0x7dc
...
Process kworker/1:2 (pid: 88, stack limit = 0x(ptrval))
Stack: (0xed9f1f00 to 0xed9f2000)
...
[<c0148590>] (process_one_work) from [<c0148d6c>] (worker_thread+0x44/0x51c)
[<c0148d6c>] (worker_thread) from [<c01500c0>] (kthread+0x158/0x1a0)
[<c01500c0>] (kthread) from [<c0100114>] (ret_from_fork+0x14/0x20)
Exception stack(0xed9f1fb0 to 0xed9f1ff8)
...
---[ end trace 5033c1326a62e5f3 ]---
note: kworker/1:2[88] exited with preempt_count 1

Fix this by flushing pending work in error path.

Reviewed-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
4 years agousb: cdns3: gadget: improve the set_configuration handling
Peter Chen [Fri, 17 Jul 2020 10:13:17 +0000 (18:13 +0800)]
usb: cdns3: gadget: improve the set_configuration handling

- Delete the duplicated EP_CMD_ERDY and EP_CMD_REQ_CMPL setting
- Prepare the next setup before setting EP_CMD_ERDY and EP_CMD_REQ_CMPL,
it could avoid a bug that DMA hang at EP0 OUT for DEV_VER_NXP_V1
- Delete the duplicated cdns3_set_hw_configuration calling at
cdns3_req_ep0_set_configuration, the composite.c will handle
this request, and call .ep0_queue back, and at .ep_queue it will
call cdns3_set_hw_configuration.
- Move cdns3_allow_enable_l1 into cdns3_set_hw_configuration since
it is part of this function.

Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
4 years agousb: dwc3: meson-g12a: fix shared reset control use
Dan Robertson [Mon, 13 Jul 2020 16:05:22 +0000 (12:05 -0400)]
usb: dwc3: meson-g12a: fix shared reset control use

The reset is a shared reset line, but reset_control_reset is still used
and reset_control_deassert is not guaranteed to have been called before
the first reset_control_assert call. When suspending the following
warning may be seen:

WARNING: CPU: 1 PID: 5530 at drivers/reset/core.c:355 reset_control_assert+0x184/0x19c
Hardware name: Hardkernel ODROID-N2 (DT)
[..]
pc : reset_control_assert+0x184/0x19c
lr : dwc3_meson_g12a_suspend+0x68/0x7c
[..]
Call trace:
 reset_control_assert+0x184/0x19c
 dwc3_meson_g12a_suspend+0x68/0x7c
 platform_pm_suspend+0x28/0x54
 __device_suspend+0x590/0xabc
 dpm_suspend+0x104/0x404
 dpm_suspend_start+0x84/0x1bc
 suspend_devices_and_enter+0xc4/0x4fc
 pm_suspend+0x198/0x2d4

Fixes: 6d9fa35a347a87 ("usb: dwc3: meson-g12a: get the reset as shared")
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
4 years agouprobes: Change handle_swbp() to send SIGTRAP with si_code=SI_KERNEL, to fix GDB...
Oleg Nesterov [Thu, 23 Jul 2020 15:44:20 +0000 (17:44 +0200)]
uprobes: Change handle_swbp() to send SIGTRAP with si_code=SI_KERNEL, to fix GDB regression

If a tracee is uprobed and it hits int3 inserted by debugger, handle_swbp()
does send_sig(SIGTRAP, current, 0) which means si_code == SI_USER. This used
to work when this code was written, but then GDB started to validate si_code
and now it simply can't use breakpoints if the tracee has an active uprobe:

# cat test.c
void unused_func(void)
{
}
int main(void)
{
return 0;
}

# gcc -g test.c -o test
# perf probe -x ./test -a unused_func
# perf record -e probe_test:unused_func gdb ./test -ex run
GNU gdb (GDB) 10.0.50.20200714-git
...
Program received signal SIGTRAP, Trace/breakpoint trap.
0x00007ffff7ddf909 in dl_main () from /lib64/ld-linux-x86-64.so.2
(gdb)

The tracee hits the internal breakpoint inserted by GDB to monitor shared
library events but GDB misinterprets this SIGTRAP and reports a signal.

Change handle_swbp() to use force_sig(SIGTRAP), this matches do_int3_user()
and fixes the problem.

This is the minimal fix for -stable, arch/x86/kernel/uprobes.c is equally
wrong; it should use send_sigtrap(TRAP_TRACE) instead of send_sig(SIGTRAP),
but this doesn't confuse GDB and needs another x86-specific patch.

Reported-by: Aaron Merey <amerey@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200723154420.GA32043@redhat.com
4 years agosched: Warn if garbage is passed to default_wake_function()
Chris Wilson [Thu, 23 Jul 2020 20:10:42 +0000 (21:10 +0100)]
sched: Warn if garbage is passed to default_wake_function()

Since the default_wake_function() passes its flags onto
try_to_wake_up(), warn if those flags collide with internal values.

Given that the supplied flags are garbage, no repair can be done but at
least alert the user to the damage they are causing.

In the belief that these errors should be picked up during testing, the
warning is only compiled in under CONFIG_SCHED_DEBUG.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: https://lore.kernel.org/r/20200723201042.18861-1-chris@chris-wilson.co.uk
4 years agoserial: exar: Fix GPIO configuration for Sealevel cards based on XR17V35X
Matthew Howell [Wed, 22 Jul 2020 20:11:24 +0000 (16:11 -0400)]
serial: exar: Fix GPIO configuration for Sealevel cards based on XR17V35X

Sealevel XR17V35X based devices are inoperable on kernel versions
4.11 and above due to a change in the GPIO preconfiguration introduced in
commit
7dea8165f1d. This patch fixes this by preconfiguring the GPIO on Sealevel
cards to the value (0x00) used prior to commit 7dea8165f1d

With GPIOs preconfigured as per commit 7dea8165f1d all ports on
Sealevel XR17V35X based devices become stuck in high impedance
mode, regardless of dip-switch or software configuration. This
causes the device to become effectively unusable. This patch (in
various forms) has been distributed to our customers and no issues
related to it have been reported.

Fixes: 7dea8165f1d6 ("serial: exar: Preconfigure xr17v35x MPIOs as output")
Signed-off-by: Matthew Howell <matthew.howell@sealevel.com>
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2007221605270.13247@tstest-VirtualBox
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Fri, 24 Jul 2020 00:22:09 +0000 (17:22 -0700)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter/IPVS fixes for net

The following patchset contains Netfilter/IPVS fixes for net:

1) Fix NAT hook deletion when table is dormant, from Florian Westphal.

2) Fix IPVS sync stalls, from guodeqing.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agogeneve: fix an uninitialized value in geneve_changelink()
Cong Wang [Thu, 23 Jul 2020 01:56:25 +0000 (18:56 -0700)]
geneve: fix an uninitialized value in geneve_changelink()

geneve_nl2info() sets 'df' conditionally, so we have to
initialize it by copying the value from existing geneve
device in geneve_changelink().

Fixes: 56c09de347e4 ("geneve: allow changing DF behavior after creation")
Reported-by: syzbot+7ebc2e088af5e4c0c9fa@syzkaller.appspotmail.com
Cc: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobonding: check return value of register_netdevice() in bond_newlink()
Cong Wang [Wed, 22 Jul 2020 23:31:54 +0000 (16:31 -0700)]
bonding: check return value of register_netdevice() in bond_newlink()

Very similar to commit 544f287b8495
("bonding: check error value of register_netdevice() immediately"),
we should immediately check the return value of register_netdevice()
before doing anything else.

Fixes: 005db31d5f5f ("bonding: set carrier off for devices created through netlink")
Reported-and-tested-by: syzbot+bbc3a11c4da63c1b74d6@syzkaller.appspotmail.com
Cc: Beniamino Galvani <bgalvani@redhat.com>
Cc: Taehee Yoo <ap420073@gmail.com>
Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoRevert "cifs: Fix the target file was deleted when rename failed."
Steve French [Thu, 23 Jul 2020 19:41:29 +0000 (14:41 -0500)]
Revert "cifs: Fix the target file was deleted when rename failed."

This reverts commit 9ffad9263b467efd8f8dc7ae1941a0a655a2bab2.

Upon additional testing with older servers, it was found that
the original commit introduced a regression when using the old SMB1
dialect and rsyncing over an existing file.

The patch will need to be respun to address this, likely including
a larger refactoring of the SMB1 and SMB3 rename code paths to make
it less confusing and also to address some additional rename error
cases that SMB3 may be able to workaround.

Signed-off-by: Steve French <stfrench@microsoft.com>
Reported-by: Patrick Fernie <patrick.fernie@gmail.com>
CC: Stable <stable@vger.kernel.org>
Acked-by: Ronnie Sahlberg <lsahlber@redhat.com>
Acked-by: Pavel Shilovsky <pshilov@microsoft.com>
Acked-by: Zhang Xiaoxu <zhangxiaoxu5@huawei.com>
4 years agoMerge tag 's390-5.8-6' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux...
Linus Torvalds [Thu, 23 Jul 2020 20:42:46 +0000 (13:42 -0700)]
Merge tag 's390-5.8-6' of git://git./linux/kernel/git/s390/linux into master

Pull s390 fixes from Heiko Carstens:

 - Change cpum_cf/perf counter name from DFLT_CCERROR to DFLT_CCFINISH
   to reflect reality and avoid further confusion. This is a user space
   visible change therefore the commit has also a stable tag for 5.7,
   where this counter was introduced.

 - Add Matthew Rosato as s390 IOMMU maintainer.

* tag 's390-5.8-6' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  MAINTAINERS: add Matthew for s390 IOMMU
  s390/cpum_cf,perf: change DFLT_CCERROR counter name

4 years agoi2c: i2c-qcom-geni: Fix DMA transfer race
Douglas Anderson [Wed, 22 Jul 2020 22:00:21 +0000 (15:00 -0700)]
i2c: i2c-qcom-geni: Fix DMA transfer race

When I have KASAN enabled on my kernel and I start stressing the
touchscreen my system tends to hang.  The touchscreen is one of the
only things that does a lot of big i2c transfers and ends up hitting
the DMA paths in the geni i2c driver.  It appears that KASAN adds
enough delay in my system to tickle a race condition in the DMA setup
code.

When the system hangs, I found that it was running the geni_i2c_irq()
over and over again.  It had these:

m_stat   = 0x04000080
rx_st    = 0x30000011
dm_tx_st = 0x00000000
dm_rx_st = 0x00000000
dma      = 0x00000001

Notably we're in DMA mode but are getting M_RX_IRQ_EN and
M_RX_FIFO_WATERMARK_EN over and over again.

Putting some traces in geni_i2c_rx_one_msg() showed that when we
failed we were getting to the start of geni_i2c_rx_one_msg() but were
never executing geni_se_rx_dma_prep().

I believe that the problem here is that we are starting the geni
command before we run geni_se_rx_dma_prep().  If a transfer makes it
far enough before we do that then we get into the state I have
observed.  Let's change the order, which seems to work fine.

Although problems were seen on the RX path, code inspection suggests
that the TX should be changed too.  Change it as well.

Fixes: 37692de5d523 ("i2c: i2c-qcom-geni: Add bus driver for the Qualcomm GENI I2C controller")
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Tested-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Reviewed-by: Akash Asthana <akashast@codeaurora.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Mukesh Kumar Savaliya <msavaliy@codeaurora.org>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
4 years agoi2c: rcar: always clear ICSAR to avoid side effects
Wolfram Sang [Sat, 4 Jul 2020 13:38:29 +0000 (15:38 +0200)]
i2c: rcar: always clear ICSAR to avoid side effects

On R-Car Gen2, we get a timeout when reading from the address set in
ICSAR, even though the slave interface is disabled. Clearing it fixes
this situation. Note that Gen3 is not affected.

To reproduce: bind and undbind an I2C slave on some bus, run
'i2cdetect' on that bus.

Fixes: de20d1857dd6 ("i2c: rcar: add slave support")
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
4 years agotcp: allow at most one TLP probe per flight
Yuchung Cheng [Thu, 23 Jul 2020 19:00:06 +0000 (12:00 -0700)]
tcp: allow at most one TLP probe per flight

Previously TLP may send multiple probes of new data in one
flight. This happens when the sender is cwnd limited. After the
initial TLP containing new data is sent, the sender receives another
ACK that acks partial inflight.  It may re-arm another TLP timer
to send more, if no further ACK returns before the next TLP timeout
(PTO) expires. The sender may send in theory a large amount of TLP
until send queue is depleted. This only happens if the sender sees
such irregular uncommon ACK pattern. But it is generally undesirable
behavior during congestion especially.

The original TLP design restrict only one TLP probe per inflight as
published in "Reducing Web Latency: the Virtue of Gentle Aggression",
SIGCOMM 2013. This patch changes TLP to send at most one probe
per inflight.

Note that if the sender is app-limited, TLP retransmits old data
and did not have this issue.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoAX.25: Prevent integer overflows in connect and sendmsg
Dan Carpenter [Thu, 23 Jul 2020 14:49:57 +0000 (17:49 +0300)]
AX.25: Prevent integer overflows in connect and sendmsg

We recently added some bounds checking in ax25_connect() and
ax25_sendmsg() and we so we removed the AX25_MAX_DIGIS checks because
they were no longer required.

Unfortunately, I believe they are required to prevent integer overflows
so I have added them back.

Fixes: 8885bb0621f0 ("AX.25: Prevent out-of-bounds read in ax25_sendmsg()")
Fixes: 2f2a7ffad5c6 ("AX.25: Fix out-of-bounds read in ax25_connect()")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodm integrity: fix integrity recalculation that is improperly skipped
Mikulas Patocka [Thu, 23 Jul 2020 14:42:09 +0000 (10:42 -0400)]
dm integrity: fix integrity recalculation that is improperly skipped

Commit adc0daad366b62ca1bce3e2958a40b0b71a8b8b3 ("dm: report suspended
device during destroy") broke integrity recalculation.

The problem is dm_suspended() returns true not only during suspend,
but also during resume. So this race condition could occur:
1. dm_integrity_resume calls queue_work(ic->recalc_wq, &ic->recalc_work)
2. integrity_recalc (&ic->recalc_work) preempts the current thread
3. integrity_recalc calls if (unlikely(dm_suspended(ic->ti))) goto unlock_ret;
4. integrity_recalc exits and no recalculating is done.

To fix this race condition, add a function dm_post_suspending that is
only true during the postsuspend phase and use it instead of
dm_suspended().

Signed-off-by: Mikulas Patocka <mpatocka redhat com>
Fixes: adc0daad366b ("dm: report suspended device during destroy")
Cc: stable vger kernel org # v4.18+
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
4 years agoio_uring: missed req_init_async() for IOSQE_ASYNC
Pavel Begunkov [Thu, 23 Jul 2020 17:17:20 +0000 (20:17 +0300)]
io_uring: missed req_init_async() for IOSQE_ASYNC

IOSQE_ASYNC branch of io_queue_sqe() is another place where an
unitialised req->work can be accessed (i.e. prior io_req_init_async()).
Nothing really bad though, it just looses IO_WQ_WORK_CONCURRENT flag.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoxhci: dbc: remove tty specific port structure from struct xhci_dbc
Mathias Nyman [Thu, 23 Jul 2020 14:45:30 +0000 (17:45 +0300)]
xhci: dbc: remove tty specific port structure from struct xhci_dbc

Use a void pointer that any function driver can use instead.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-28-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbgcap: remove dbc dependency on dbctty specific flag
Mathias Nyman [Thu, 23 Jul 2020 14:45:29 +0000 (17:45 +0300)]
xhci: dbgcap: remove dbc dependency on dbctty specific flag

dbc should not be aware of, or use any dbctty specific variables.
currenly dbc driver reads the port->registered flag to see if the
callbacks should be called.

Only makes these decisions based on dbc internal state instead.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-27-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Add a operations structure to access driver functions
Mathias Nyman [Thu, 23 Jul 2020 14:45:28 +0000 (17:45 +0300)]
xhci: dbc: Add a operations structure to access driver functions

Don't call dbctty driver functions directly from dbc core code.
Introduce a new dbc_driver structure that contains function pointers
for disconnect and configure operations.

The driver (ttydbc) must provide these opeations when creating a dbc.

Name the structure dbc_driver instead of dbc_ops as we plan to
add more driver configureable values here, such as vid and pid.

Decouples dbc and dbctty.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-26-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbctty: split dbc tty driver registration and unregistration functions.
Mathias Nyman [Thu, 23 Jul 2020 14:45:27 +0000 (17:45 +0300)]
xhci: dbctty: split dbc tty driver registration and unregistration functions.

Split the dbc tty driver registrations function into separate
init and probe parts.

The init part will register the tty driver, and should in the future be
called from module_init().
The probe part will become the normal probe function, but for now it is
called from the init part.

The unregister function is s likewise split into remove and exit parts.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-25-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: remove endpoint pointers from dbc_port structure
Mathias Nyman [Thu, 23 Jul 2020 14:45:26 +0000 (17:45 +0300)]
xhci: dbc: remove endpoint pointers from dbc_port structure

dbctty no longer needs references directly to dbc endpoints,
so remove them

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-24-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: simplify dbc requests allocation and queueing
Mathias Nyman [Thu, 23 Jul 2020 14:45:25 +0000 (17:45 +0300)]
xhci: dbc: simplify dbc requests allocation and queueing

Don't pass endpoint pointer, dbctty should not be aware of
struct dbc_ep, knowing the direction is enough.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-23-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Pass dbc pointer to dbc start and stop functions.
Mathias Nyman [Thu, 23 Jul 2020 14:45:24 +0000 (17:45 +0300)]
xhci: dbc: Pass dbc pointer to dbc start and stop functions.

xhci_dbc_start() and xhci_dbc_stop() functions only used xhci_hcd pointer
to get the dbc pointer.
Pass the dbc pointer instead of the xhci_hcd pointer as a parameter

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-22-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Pass dbc pointer to dbc memory init and cleanup functions
Mathias Nyman [Thu, 23 Jul 2020 14:45:23 +0000 (17:45 +0300)]
xhci: dbc: Pass dbc pointer to dbc memory init and cleanup functions

Dbc mem_init and mem_cleanup functions used xhci_hcd to get to the device
pointer. The device pointer can be accessed directly from dbc structure,
so pass a pointer to dbc as a parameter instead.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-21-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: don't use generic xhci ring allocation functions for dbc.
Mathias Nyman [Thu, 23 Jul 2020 14:45:22 +0000 (17:45 +0300)]
xhci: dbc: don't use generic xhci ring allocation functions for dbc.

The generic xhci ring allocations code needs struct xhci_hcd pointer, and
it allocates memory for the rings from dma pools created for the xhci
device.

In order to decouple xhci and DbC we have to create our own ring allocation
and free routines for DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-20-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Don't use generic xhci context allocation for dbc
Mathias Nyman [Thu, 23 Jul 2020 14:45:21 +0000 (17:45 +0300)]
xhci: dbc: Don't use generic xhci context allocation for dbc

The DbC context is different from the xhci device context.
It's a lot smaller as it only contains three 64 bytes sub-contexts;
the info, endpoint-out, and endpoint-in contexts. In total 192 bytes.
The context size (CSZ) field in HCCPARAMS1 xhci register does not alter
DbC context size like it does for xhci device contexts.

So don't use the geneic xhci context memory allocation, or the
dma pool that is intended for xhci device contexts.

In addition to saving memory this also helps decoupleing xhci and dbc code.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-19-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Use dbc structure in the request completion instead of xhci_hcd
Mathias Nyman [Thu, 23 Jul 2020 14:45:20 +0000 (17:45 +0300)]
xhci: dbc: Use dbc structure in the request completion instead of xhci_hcd

The dbc request completion callback doesn't need a xhci_hcd pointer.
The only user of the xhci_hcd pointer in dbgtty request callback was
the xhci_warn() function. Change it to dev_warn() instead.

While changing the callback function parameter to dbc in struct
xhci_requeset, move the struct xhci_request declaraion down a bit in the
header file to avoid compiler warinings

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-18-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Pass dbc pointer to get_in/out_ep() helper functions to get endpoints
Mathias Nyman [Thu, 23 Jul 2020 14:45:19 +0000 (17:45 +0300)]
xhci: dbc: Pass dbc pointer to get_in/out_ep() helper functions to get endpoints

Pass dbc pointer instead of struct xhci_hcd pointer to the get_in_ep() and
get_out_ep() helper functions.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-17-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbgtty: Pass dbc pointer when registering a dbctty device
Mathias Nyman [Thu, 23 Jul 2020 14:45:18 +0000 (17:45 +0300)]
xhci: dbgtty: Pass dbc pointer when registering a dbctty device

Pass dbc pointer to the xhci_dbc_tty_register_device() and
xhci_dbc_tty_unregister_device() functions instead of xhci_hcd pointer

These functions don't need a xhci_hcd pointer anymore, only use case was
the xhci_err() function, which is now changed to a dev_err() instead.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-16-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Pass dbc pointer to dbc_handle_xfer_event() instead of xhci_hcd pointer
Mathias Nyman [Thu, 23 Jul 2020 14:45:17 +0000 (17:45 +0300)]
xhci: dbc: Pass dbc pointer to dbc_handle_xfer_event() instead of xhci_hcd pointer

The event handling function only used xhci pointer to get the dbc pointer.
Pass the dbc pointer instead as a parameter

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-15-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Change to pass dbc pointer to xhci_do_dbc_stop()
Mathias Nyman [Thu, 23 Jul 2020 14:45:16 +0000 (17:45 +0300)]
xhci: dbc: Change to pass dbc pointer to xhci_do_dbc_stop()

Pass the dbc pointer instead of xhci_hcd pointer in order to
decouple xhci and dbc.

xhci_do_dbc_stop() only used xhci to get the dbc pointer.
Pass the dbc pointer instead as a parameter

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-14-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Pass dbc pointer to endpoint init and exit functions.
Mathias Nyman [Thu, 23 Jul 2020 14:45:15 +0000 (17:45 +0300)]
xhci: dbc: Pass dbc pointer to endpoint init and exit functions.

struct xhci_hcd pointer is not needed for dbc endpoint init and exit,
it was only used to get to the dbc structure.
Pass the dbc pointer as a parameter to these functions instead.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-13-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Get the device pointer from dbc structure in dbc_ep_do_queue()
Mathias Nyman [Thu, 23 Jul 2020 14:45:14 +0000 (17:45 +0300)]
xhci: dbc: Get the device pointer from dbc structure in dbc_ep_do_queue()

dbc_ep_do_queue() can now get the device pointer directly from dbc
structure instead of going through the xhci_hcd structure.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-12-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Don't pass the xhci pointer as a parameter to xhci_dbc_init_context()
Mathias Nyman [Thu, 23 Jul 2020 14:45:13 +0000 (17:45 +0300)]
xhci: dbc: Don't pass the xhci pointer as a parameter to xhci_dbc_init_context()

xhci_dbc_init_context() no longer needs the struct xhci_hcd pointer.
Pass the dbc pointer directly instead.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-11-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Don't use xhci_write_64() as it takes xhci as a parameter
Mathias Nyman [Thu, 23 Jul 2020 14:45:12 +0000 (17:45 +0300)]
xhci: dbc: Don't use xhci_write_64() as it takes xhci as a parameter

xhci_write_64() is essentially a wrapper for lo_hi_writeq(), but it
requires struct xhci_hcd * as a parameter.
Use lo_hi_writeq() directly instead

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-10-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Use dev_info() and similar instead of xhci_info()
Mathias Nyman [Thu, 23 Jul 2020 14:45:11 +0000 (17:45 +0300)]
xhci: dbc: Use dev_info() and similar instead of xhci_info()

To make this change possible we also need to change
dbc_handle_port_status() to take dbc pointer as parameter instead of
xhci_hcd pointer.

Note that xhci_info() used xhci_to_hcd(xhci)->self.controller as the
device while for dev_info we use xhci_to_hcd(xhci)->self.sysdev.

In many cases those are the same, but not for some device where
a dwc3 controller creates a xhci platform device. In th this case
self.controller may be the platform device while self.sysdev is
the actual device known to firmware (dwc3).

This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-9-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Add device pointer to dbc structure
Mathias Nyman [Thu, 23 Jul 2020 14:45:10 +0000 (17:45 +0300)]
xhci: dbc: Add device pointer to dbc structure

Currently the dbc structure contains a pointer to struct xhci_hcd,
and dbc functions use that to dig up the underlying device pointer.

We are trying to decouple xhci and dbc code, and prepare for
code that use dbc such as dbctty into into real device drivers.
This is one step along the way.

Keep functionality the same and keep the xhci pointer, and
let the new device pointer point to the xhci device for now.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-8-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Remove dbc_dma_free_coherent() wrapper
Mathias Nyman [Thu, 23 Jul 2020 14:45:09 +0000 (17:45 +0300)]
xhci: dbc: Remove dbc_dma_free_coherent() wrapper

dbc_dma_free_coherent() takes struct xhci_hcd pointer as a parameter, but
does nothing more than calls dma_free_coherent().
Remove it and call dma_free_coherent() directly instead.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-7-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Remove dbc_dma_alloc_coherent() wrapper
Mathias Nyman [Thu, 23 Jul 2020 14:45:08 +0000 (17:45 +0300)]
xhci: dbc: Remove dbc_dma_alloc_coherent() wrapper

dbc_dma_alloc_coherent() takes struct xhci_hcd pointer as an parameter, but
does nothing more than calls dma_alloc_coherent().
Remove it and call dma_alloc_coherent() directly instead.

No functional changes
This change helps decoupling xhci and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-6-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Don't use generic xhci erst allocation and free functions
Mathias Nyman [Thu, 23 Jul 2020 14:45:07 +0000 (17:45 +0300)]
xhci: dbc: Don't use generic xhci erst allocation and free functions

The generic erst allocation and free functions take struct xhci_hcd pointer
as a parameter. Create own erst helpers for DbC in order to decouple xhci
and DbC

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-5-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: Don't pass struct xhci_hcd pointer to xhci_link_seg()
Mathias Nyman [Thu, 23 Jul 2020 14:45:06 +0000 (17:45 +0300)]
xhci: Don't pass struct xhci_hcd pointer to xhci_link_seg()

It's only used to dig out if we need to set a chain flag for specific
hosts. Pass the flag directly as a parameter instead.

No functional changes.

xhci_link_seg() is also used by DbC code, this change helps decoupling
xhci and DbC.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-4-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: dbc: Don't use generic xhci inc_deq() function for dbc
Mathias Nyman [Thu, 23 Jul 2020 14:45:05 +0000 (17:45 +0300)]
xhci: dbc: Don't use generic xhci inc_deq() function for dbc

The generic inc_deq() helper takes struct xhci_hcd pointer as a parameter,
and is a lot more complex than needed for the DbC usecase.

In order to decouple xhci and DbC we have to create our own small
inc_evt_deq() helper, not relying on xhci.

Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-3-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxhci: Make debug message consistent with bus and port number
Kai-Heng Feng [Thu, 23 Jul 2020 14:45:04 +0000 (17:45 +0300)]
xhci: Make debug message consistent with bus and port number

Current xhci debug message doesn't always output bus number, so it's
hard to figure out it's from USB2 or USB3 root hub.

In addition to that, some port numbers are offset to 0 and others are
offset to 1. Use the latter to match the USB core.

So use "bus number - port index + 1" to make debug message consistent.

Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Link: https://lore.kernel.org/r/20200723144530.9992-2-mathias.nyman@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agodevice property: Avoid NULL pointer dereference in device_get_next_child_node()
Andy Shevchenko [Thu, 16 Jul 2020 18:27:47 +0000 (21:27 +0300)]
device property: Avoid NULL pointer dereference in device_get_next_child_node()

When we have no primary fwnode or when it's a software node, we may end up
in the situation when fwnode is a NULL pointer. There is no point to look for
secondary fwnode in such case. Add a necessary check to a condition.

Fixes: 114dbb4fa7c4 ("drivers property: When no children in primary, try secondary")
Reported-by: Maxim Levitsky <mlevitsk@redhat.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Tested-by: Maxim Levitsky <mlevitsk@redhat.com>
Link: https://lore.kernel.org/r/20200716182747.54929-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agofbdev: Detect integer underflow at "struct fbcon_ops"->clear_margins.
Tetsuo Handa [Wed, 15 Jul 2020 01:51:02 +0000 (10:51 +0900)]
fbdev: Detect integer underflow at "struct fbcon_ops"->clear_margins.

syzbot is reporting general protection fault in bitfill_aligned() [1]
caused by integer underflow in bit_clear_margins(). The cause of this
problem is when and how do_vc_resize() updates vc->vc_{cols,rows}.

If vc_do_resize() fails (e.g. kzalloc() fails) when var.xres or var.yres
is going to shrink, vc->vc_{cols,rows} will not be updated. This allows
bit_clear_margins() to see info->var.xres < (vc->vc_cols * cw) or
info->var.yres < (vc->vc_rows * ch). Unexpectedly large rw or bh will
try to overrun the __iomem region and causes general protection fault.

Also, vc_resize(vc, 0, 0) does not set vc->vc_{cols,rows} = 0 due to

  new_cols = (cols ? cols : vc->vc_cols);
  new_rows = (lines ? lines : vc->vc_rows);

exception. Since cols and lines are calculated as

  cols = FBCON_SWAP(ops->rotate, info->var.xres, info->var.yres);
  rows = FBCON_SWAP(ops->rotate, info->var.yres, info->var.xres);
  cols /= vc->vc_font.width;
  rows /= vc->vc_font.height;
  vc_resize(vc, cols, rows);

in fbcon_modechanged(), var.xres < vc->vc_font.width makes cols = 0
and var.yres < vc->vc_font.height makes rows = 0. This means that

  const int fd = open("/dev/fb0", O_ACCMODE);
  struct fb_var_screeninfo var = { };
  ioctl(fd, FBIOGET_VSCREENINFO, &var);
  var.xres = var.yres = 1;
  ioctl(fd, FBIOPUT_VSCREENINFO, &var);

easily reproduces integer underflow bug explained above.

Of course, callers of vc_resize() are not handling vc_do_resize() failure
is bad. But we can't avoid vc_resize(vc, 0, 0) which returns 0. Therefore,
as a band-aid workaround, this patch checks integer underflow in
"struct fbcon_ops"->clear_margins call, assuming that
vc->vc_cols * vc->vc_font.width and vc->vc_rows * vc->vc_font.heigh do not
cause integer overflow.

[1] https://syzkaller.appspot.com/bug?id=a565882df74fa76f10d3a6fec4be31098dbb37c6

Reported-and-tested-by: syzbot <syzbot+e5fd3e65515b48c02a30@syzkaller.appspotmail.com>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200715015102.3814-1-penguin-kernel@I-love.SAKURA.ne.jp
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agousb: dwc2: don't use ID/Vbus detection if usb-role-switch on STM32MP15 SoCs
Amelie Delaunay [Tue, 16 Jun 2020 14:07:16 +0000 (16:07 +0200)]
usb: dwc2: don't use ID/Vbus detection if usb-role-switch on STM32MP15 SoCs

If usb-role-switch is present in the device tree, it means that ID and Vbus
signals are not connected to the OTG controller but to an external
component (GPIOs, Type-C controller). In this configuration, usb role
switch is used to force valid sessions on STM32MP15 SoCs.

Acked-by: Minas Harutyunyan <hminas@synopsys.com>
Signed-off-by: Amelie Delaunay <amelie.delaunay@st.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
4 years agousb: dwc2: override PHY input signals with usb role switch support
Amelie Delaunay [Tue, 16 Jun 2020 14:07:15 +0000 (16:07 +0200)]
usb: dwc2: override PHY input signals with usb role switch support

This patch adds support for usb role switch to dwc2, by using overriding
control of the PHY voltage valid and ID input signals.

iddig signal (ID) can be overridden:
- when setting GUSBCFG_FORCEHOSTMODE, iddig input pin is overridden with 1;
- when setting GUSBCFG_FORCEDEVMODE, iddig input pin is overridden with 0.

avalid/bvalid/vbusvalid signals can be overridden respectively with:
- GOTGCTL_AVALOEN + GOTGCTL_AVALOVAL
- GOTGCTL_BVALOEN + GOTGCTL_BVALOVAL
- GOTGCTL_VBVALEN + GOTGCTL_VBVALOVAL

It is possible to determine valid sessions thanks to usb role switch:
- if USB_ROLE_NONE then !avalid && !bvalid && !vbusvalid
- if USB_ROLE_DEVICE then !avalid && bvalid && vbusvalid
- if USB_ROLE_HOST then avalid && !bvalid && vbusvalid

Acked-by: Minas Harutyunyan <hminas@synopsys.com>
Signed-off-by: Amelie Delaunay <amelie.delaunay@st.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
4 years agousb: tegra: Fix allocation for the FPCI context
Jon Hunter [Wed, 15 Jul 2020 11:38:42 +0000 (12:38 +0100)]
usb: tegra: Fix allocation for the FPCI context

Commit 5c4e8d3781bc ("usb: host: xhci-tegra: Add support for XUSB
context save/restore") is using the IPFS 'num_offsets' value when
allocating memory for FPCI context instead of the FPCI 'num_offsets'.

After commit cad064f1bd52 ("devres: handle zero size in devm_kmalloc()")
was added system suspend started failing on Tegra186. The kernel log
showed that the Tegra XHCI driver was crashing on entry to suspend when
attempting the save the USB context. On Tegra186, the IPFS context has a
zero length but the FPCI content has a non-zero length, and because of
the bug in the Tegra XHCI driver we are incorrectly allocating a zero
length array for the FPCI context. The crash seen on entering suspend
when we attempt to save the FPCI context and following commit
cad064f1bd52 ("devres: handle zero size in devm_kmalloc()") this now
causes a NULL pointer deference when we access the memory. Fix this by
correcting the amount of memory we are allocating for FPCI contexts.

Cc: stable@vger.kernel.org
Fixes: 5c4e8d3781bc ("usb: host: xhci-tegra: Add support for XUSB context save/restore")

Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20200715113842.30680-1-jonathanh@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoarm64: vdso32: Fix '--prefix=' value for newer versions of clang
Nathan Chancellor [Thu, 23 Jul 2020 04:15:10 +0000 (21:15 -0700)]
arm64: vdso32: Fix '--prefix=' value for newer versions of clang

Newer versions of clang only look for $(COMPAT_GCC_TOOLCHAIN_DIR)as [1],
rather than $(COMPAT_GCC_TOOLCHAIN_DIR)$(CROSS_COMPILE_COMPAT)as,
resulting in the following build error:

$ make -skj"$(nproc)" ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- \
CROSS_COMPILE_COMPAT=arm-linux-gnueabi- LLVM=1 O=out/aarch64 distclean \
defconfig arch/arm64/kernel/vdso32/
...
/home/nathan/cbl/toolchains/llvm-binutils/bin/as: unrecognized option '-EL'
clang-12: error: assembler command failed with exit code 1 (use -v to see invocation)
make[3]: *** [arch/arm64/kernel/vdso32/Makefile:181: arch/arm64/kernel/vdso32/note.o] Error 1
...

Adding the value of CROSS_COMPILE_COMPAT (adding notdir to account for a
full path for CROSS_COMPILE_COMPAT) fixes this issue, which matches the
solution done for the main Makefile [2].

[1]: https://github.com/llvm/llvm-project/commit/3452a0d8c17f7166f479706b293caf6ac76ffd90
[2]: https://lore.kernel.org/lkml/20200721173125.1273884-1-maskray@google.com/

Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Cc: stable@vger.kernel.org
Link: https://github.com/ClangBuiltLinux/linux/issues/1099
Link: https://lore.kernel.org/r/20200723041509.400450-1-natechancellor@gmail.com
Signed-off-by: Will Deacon <will@kernel.org>
4 years agointerconnect: msm8916: Fix buswidth of pcnoc_s nodes
Georgi Djakov [Thu, 23 Jul 2020 08:37:35 +0000 (11:37 +0300)]
interconnect: msm8916: Fix buswidth of pcnoc_s nodes

The buswidth of the pcnoc_s_* nodes is actually not 8, but
4 bytes. Let's fix it.

Reported-by: Jun Nie <jun.nie@linaro.org>
Reviewed-by: Mike Tipton <mdtipton@codeaurora.org>
Fixes: 30c8fa3ec61a ("interconnect: qcom: Add MSM8916 interconnect provider driver")
Link: https://lore.kernel.org/r/20200709130004.12462-1-georgi.djakov@linaro.org
Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200723083735.5616-3-georgi.djakov@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agointerconnect: Do not skip aggregation for disabled paths
Georgi Djakov [Thu, 23 Jul 2020 08:37:34 +0000 (11:37 +0300)]
interconnect: Do not skip aggregation for disabled paths

When an interconnect path is being disabled, currently we don't aggregate
the requests for it afterwards. But the re-aggregation step shouldn't be
skipped, as it may leave the nodes with outdated bandwidth data. This
outdated data may actually keep the path still enabled and prevent the
device from going into lower power states.

Reported-by: Atul Dhudase <adhudase@codeaurora.org>
Fixes: 7d374b209083 ("interconnect: Add helpers for enabling/disabling a path")
Reviewed-by: Sibi Sankar <sibis@codeaurora.org>
Tested-by: Atul Dhudase <adhudase@codeaurora.org>
Reviewed-by: Atul Dhudase <adhudase@codeaurora.org>
Link: https://lore.kernel.org/r/20200721120740.3436-1-georgi.djakov@linaro.org
Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Link: https://lore.kernel.org/r/20200723083735.5616-2-georgi.djakov@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years ago/dev/mem: Add missing memory barriers for devmem_inode
Eric Biggers [Thu, 16 Jul 2020 06:05:53 +0000 (23:05 -0700)]
/dev/mem: Add missing memory barriers for devmem_inode

WRITE_ONCE() isn't the correct way to publish a pointer to a data
structure, since it doesn't include a write memory barrier.  Therefore
other tasks may see that the pointer has been set but not see that the
pointed-to memory has finished being initialized yet.  Instead a
primitive with "release" semantics is needed.

Use smp_store_release() for this.

The use of READ_ONCE() on the read side is still potentially correct if
there's no control dependency, i.e. if all memory being "published" is
transitively reachable via the pointer itself.  But this pairing is
somewhat confusing and error-prone.  So just upgrade the read side to
smp_load_acquire() so that it clearly pairs with smp_store_release().

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Fixes: 3234ac664a87 ("/dev/mem: Revoke mappings when a driver claims the region")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: stable <stable@vger.kernel.org>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/20200716060553.24618-1-ebiggers@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobinder: Don't use mmput() from shrinker function.
Tetsuo Handa [Thu, 16 Jul 2020 15:12:15 +0000 (00:12 +0900)]
binder: Don't use mmput() from shrinker function.

syzbot is reporting that mmput() from shrinker function has a risk of
deadlock [1], for delayed_uprobe_add() from update_ref_ctr() calls
kzalloc(GFP_KERNEL) with delayed_uprobe_lock held, and
uprobe_clear_state() from __mmput() also holds delayed_uprobe_lock.

Commit a1b2289cef92ef0e ("android: binder: drop lru lock in isolate
callback") replaced mmput() with mmput_async() in order to avoid sleeping
with spinlock held. But this patch replaces mmput() with mmput_async() in
order not to start __mmput() from shrinker context.

[1] https://syzkaller.appspot.com/bug?id=bc9e7303f537c41b2b0cc2dfcea3fc42964c2d45

Reported-by: syzbot <syzbot+1068f09c44d151250c33@syzkaller.appspotmail.com>
Reported-by: syzbot <syzbot+e5344baa319c9a96edec@syzkaller.appspotmail.com>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Acked-by: Todd Kjos <tkjos@google.com>
Acked-by: Christian Brauner <christian.brauner@ubuntu.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/4ba9adb2-43f5-2de0-22de-f6075c1fab50@i-love.sakura.ne.jp
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoMerge tag 'amd-drm-fixes-5.8-2020-07-22' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Thu, 23 Jul 2020 04:06:14 +0000 (14:06 +1000)]
Merge tag 'amd-drm-fixes-5.8-2020-07-22' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

amd-drm-fixes-5.8-2020-07-22:

amdgpu:
- Fix crash when overclocking VegaM
- Fix possible crash when editing dpm levels

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200723032608.3865-1-alexander.deucher@amd.com
4 years agoMerge tag 'drm-misc-fixes-2020-07-22' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Thu, 23 Jul 2020 04:05:28 +0000 (14:05 +1000)]
Merge tag 'drm-misc-fixes-2020-07-22' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

 * sun4i: Fix inverted HPD result; fixes an earlier fix
 * lima: fix timeout during reset

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20200722070321.GA29190@linux-uq9g
4 years agocxgb4: add missing release on skb in uld_send()
Navid Emamdoost [Thu, 23 Jul 2020 02:58:39 +0000 (21:58 -0500)]
cxgb4: add missing release on skb in uld_send()

In the implementation of uld_send(), the skb is consumed on all
execution paths except one. Release skb when returning NET_XMIT_DROP.

Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMakefile: Fix GCC_TOOLCHAIN_DIR prefix for Clang cross compilation
Fangrui Song [Tue, 21 Jul 2020 17:31:23 +0000 (10:31 -0700)]
Makefile: Fix GCC_TOOLCHAIN_DIR prefix for Clang cross compilation

When CROSS_COMPILE is set (e.g. aarch64-linux-gnu-), if
$(CROSS_COMPILE)elfedit is found at /usr/bin/aarch64-linux-gnu-elfedit,
GCC_TOOLCHAIN_DIR will be set to /usr/bin/.  --prefix= will be set to
/usr/bin/ and Clang as of 11 will search for both
$(prefix)aarch64-linux-gnu-$needle and $(prefix)$needle.

GCC searchs for $(prefix)aarch64-linux-gnu/$version/$needle,
$(prefix)aarch64-linux-gnu/$needle and $(prefix)$needle. In practice,
$(prefix)aarch64-linux-gnu/$needle rarely contains executables.

To better model how GCC's -B/--prefix takes in effect in practice, newer
Clang (since
https://github.com/llvm/llvm-project/commit/3452a0d8c17f7166f479706b293caf6ac76ffd90)
only searches for $(prefix)$needle. Currently it will find /usr/bin/as
instead of /usr/bin/aarch64-linux-gnu-as.

Set --prefix= to $(GCC_TOOLCHAIN_DIR)$(notdir $(CROSS_COMPILE))
(/usr/bin/aarch64-linux-gnu-) so that newer Clang can find the
appropriate cross compiling GNU as (when -no-integrated-as is in
effect).

Cc: stable@vger.kernel.org
Reported-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/1099
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
4 years agonet: atlantic: fix PTP on AQC10X
Egor Pomozov [Wed, 22 Jul 2020 19:09:58 +0000 (22:09 +0300)]
net: atlantic: fix PTP on AQC10X

This patch fixes PTP on AQC10X.
PTP support on AQC10X requires FW involvement and FW configures the
TPS data arb mode itself.
So we must make sure driver doesn't touch TPS data arb mode on AQC10x
if PTP is enabled. Otherwise, there are no timestamps even though
packets are flowing.

Fixes: 2deac71ac492a ("net: atlantic: QoS implementation: min_rate")
Signed-off-by: Egor Pomozov <epomozov@marvell.com>
Signed-off-by: Mark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoAX.25: Prevent out-of-bounds read in ax25_sendmsg()
Peilin Ye [Wed, 22 Jul 2020 16:05:12 +0000 (12:05 -0400)]
AX.25: Prevent out-of-bounds read in ax25_sendmsg()

Checks on `addr_len` and `usax->sax25_ndigis` are insufficient.
ax25_sendmsg() can go out of bounds when `usax->sax25_ndigis` equals to 7
or 8. Fix it.

It is safe to remove `usax->sax25_ndigis > AX25_MAX_DIGIS`, since
`addr_len` is guaranteed to be less than or equal to
`sizeof(struct full_sockaddr_ax25)`

Signed-off-by: Peilin Ye <yepeilin.cs@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'sctp-shrink-stream-outq-in-the-right-place'
David S. Miller [Thu, 23 Jul 2020 01:00:12 +0000 (18:00 -0700)]
Merge branch 'sctp-shrink-stream-outq-in-the-right-place'

Xin Long says:

====================
sctp: shrink stream outq in the right place

Patch 1 is an improvement, and Patch 2 is a bug fix.
====================

Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agosctp: shrink stream outq when fails to do addstream reconf
Xin Long [Wed, 22 Jul 2020 15:52:12 +0000 (23:52 +0800)]
sctp: shrink stream outq when fails to do addstream reconf

When adding a stream with stream reconf, the new stream firstly is in
CLOSED state but new out chunks can still be enqueued. Then once gets
the confirmation from the peer, the state will change to OPEN.

However, if the peer denies, it needs to roll back the stream. But when
doing that, it only sets the stream outcnt back, and the chunks already
in the new stream don't get purged. It caused these chunks can still be
dequeued in sctp_outq_dequeue_data().

As its stream is still in CLOSE, the chunk will be enqueued to the head
again by sctp_outq_head_data(). This chunk will never be sent out, and
the chunks after it can never be dequeued. The assoc will be 'hung' in
a dead loop of sending this chunk.

To fix it, this patch is to purge these chunks already in the new
stream by calling sctp_stream_shrink_out() when failing to do the
addstream reconf.

Fixes: 11ae76e67a17 ("sctp: implement receiver-side procedures for the Reconf Response Parameter")
Reported-by: Ying Xu <yinxu@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agosctp: shrink stream outq only when new outcnt < old outcnt
Xin Long [Wed, 22 Jul 2020 15:52:11 +0000 (23:52 +0800)]
sctp: shrink stream outq only when new outcnt < old outcnt

It's not necessary to go list_for_each for outq->out_chunk_list
when new outcnt >= old outcnt, as no chunk with higher sid than
new (outcnt - 1) exists in the outqueue.

While at it, also move the list_for_each code in a new function
sctp_stream_shrink_out(), which will be used in the next patch.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoAX.25: Fix out-of-bounds read in ax25_connect()
Peilin Ye [Wed, 22 Jul 2020 15:19:01 +0000 (11:19 -0400)]
AX.25: Fix out-of-bounds read in ax25_connect()

Checks on `addr_len` and `fsa->fsa_ax25.sax25_ndigis` are insufficient.
ax25_connect() can go out of bounds when `fsa->fsa_ax25.sax25_ndigis`
equals to 7 or 8. Fix it.

This issue has been reported as a KMSAN uninit-value bug, because in such
a case, ax25_connect() reaches into the uninitialized portion of the
`struct sockaddr_storage` statically allocated in __sys_connect().

It is safe to remove `fsa->fsa_ax25.sax25_ndigis > AX25_MAX_DIGIS` because
`addr_len` is guaranteed to be less than or equal to
`sizeof(struct full_sockaddr_ax25)`.

Reported-by: syzbot+c82752228ed975b0a623@syzkaller.appspotmail.com
Link: https://syzkaller.appspot.com/bug?id=55ef9d629f3b3d7d70b69558015b63b48d01af66
Signed-off-by: Peilin Ye <yepeilin.cs@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoenetc: Remove the mdio bus on PF probe bailout
Claudiu Manoil [Wed, 22 Jul 2020 14:40:12 +0000 (17:40 +0300)]
enetc: Remove the mdio bus on PF probe bailout

For ENETC ports that register an external MDIO bus,
the bus doesn't get removed on the error bailout path
of enetc_pf_probe().

This issue became much more visible after recent:
commit 07095c025ac2 ("net: enetc: Use DT protocol information to set up the ports")
Before this commit, one could make probing fail on the error
path only by having register_netdev() fail, which is unlikely.
But after this commit, because it moved the enetc_of_phy_get()
call up in the probing sequence, now we can trigger an mdiobus_free()
bug just by forcing enetc_alloc_msix() to return error, i.e. with the
'pci=nomsi' kernel bootarg (since ENETC relies on MSI support to work),
as the calltrace below shows:

kernel BUG at /home/eiz/work/enetc/net/drivers/net/phy/mdio_bus.c:648!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[...]
Hardware name: LS1028A RDB Board (DT)
pstate: 80000005 (Nzcv daif -PAN -UAO BTYPE=--)
pc : mdiobus_free+0x50/0x58
lr : devm_mdiobus_free+0x14/0x20
[...]
Call trace:
 mdiobus_free+0x50/0x58
 devm_mdiobus_free+0x14/0x20
 release_nodes+0x138/0x228
 devres_release_all+0x38/0x60
 really_probe+0x1c8/0x368
 driver_probe_device+0x5c/0xc0
 device_driver_attach+0x74/0x80
 __driver_attach+0x8c/0xd8
 bus_for_each_dev+0x7c/0xd8
 driver_attach+0x24/0x30
 bus_add_driver+0x154/0x200
 driver_register+0x64/0x120
 __pci_register_driver+0x44/0x50
 enetc_pf_driver_init+0x24/0x30
 do_one_initcall+0x60/0x1c0
 kernel_init_freeable+0x1fc/0x274
 kernel_init+0x14/0x110
 ret_from_fork+0x10/0x34

Fixes: ebfcb23d62ab ("enetc: Add ENETC PF level external MDIO support")
Signed-off-by: Claudiu Manoil <claudiu.manoil@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'efi-urgent-for-v5.8-rc4' of git://git.kernel.org/pub/scm/linux/kernel...
Thomas Gleixner [Wed, 22 Jul 2020 22:46:44 +0000 (00:46 +0200)]
Merge tag 'efi-urgent-for-v5.8-rc4' of git://git./linux/kernel/git/efi/efi into efi/urgent

Pull EFI fixes from Ard Biesheuvel:

 - Fix the layering violation in the use of the EFI runtime services
   availability mask in users of the 'efivars' abstraction
 - Revert build fix for GCC v4.8 which is no longer supported
 - Some fixes for build issues found by Atish while working on RISC-V support
 - Avoid --whole-archive when linking the stub on arm64
 - Some x86 EFI stub cleanups from Arvind

4 years agox86/dumpstack: Dump user space code correctly again
Thomas Gleixner [Wed, 22 Jul 2020 08:39:54 +0000 (10:39 +0200)]
x86/dumpstack: Dump user space code correctly again

H.J. reported that post 5.7 a segfault of a user space task does not longer
dump the Code bytes when /proc/sys/debug/exception-trace is enabled. It
prints 'Code: Bad RIP value.' instead.

This was broken by a recent change which made probe_kernel_read() reject
non-kernel addresses.

Update show_opcodes() so it retrieves user space opcodes via
copy_from_user_nmi().

Fixes: 98a23609b103 ("maccess: always use strict semantics for probe_kernel_read")
Reported-by: H.J. Lu <hjl.tools@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/87h7tz306w.fsf@nanos.tec.linutronix.de
4 years agox86/stacktrace: Fix reliable check for empty user task stacks
Josh Poimboeuf [Fri, 17 Jul 2020 14:04:26 +0000 (09:04 -0500)]
x86/stacktrace: Fix reliable check for empty user task stacks

If a user task's stack is empty, or if it only has user regs, ORC
reports it as a reliable empty stack.  But arch_stack_walk_reliable()
incorrectly treats it as unreliable.

That happens because the only success path for user tasks is inside the
loop, which only iterates on non-empty stacks.  Generally, a user task
must end in a user regs frame, but an empty stack is an exception to
that rule.

Thanks to commit 71c95825289f ("x86/unwind/orc: Fix error handling in
__unwind_start()"), unwind_start() now sets state->error appropriately.
So now for both ORC and FP unwinders, unwind_done() and !unwind_error()
always means the end of the stack was successfully reached.  So the
success path for kthreads is no longer needed -- it can also be used for
empty user tasks.

Reported-by: Wang ShaoBo <bobo.shaobowang@huawei.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Wang ShaoBo <bobo.shaobowang@huawei.com>
Link: https://lkml.kernel.org/r/f136a4e5f019219cbc4f4da33b30c2f44fa65b84.1594994374.git.jpoimboe@redhat.com
4 years agox86/unwind/orc: Fix ORC for newly forked tasks
Josh Poimboeuf [Fri, 17 Jul 2020 14:04:25 +0000 (09:04 -0500)]
x86/unwind/orc: Fix ORC for newly forked tasks

The ORC unwinder fails to unwind newly forked tasks which haven't yet
run on the CPU.  It correctly reads the 'ret_from_fork' instruction
pointer from the stack, but it incorrectly interprets that value as a
call stack address rather than a "signal" one, so the address gets
incorrectly decremented in the call to orc_find(), resulting in bad ORC
data.

Fix it by forcing 'ret_from_fork' frames to be signal frames.

Reported-by: Wang ShaoBo <bobo.shaobowang@huawei.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Wang ShaoBo <bobo.shaobowang@huawei.com>
Link: https://lkml.kernel.org/r/f91a8778dde8aae7f71884b5df2b16d552040441.1594994374.git.jpoimboe@redhat.com
4 years agonfsd4: fix NULL dereference in nfsd/clients display code
J. Bruce Fields [Wed, 15 Jul 2020 17:31:36 +0000 (13:31 -0400)]
nfsd4: fix NULL dereference in nfsd/clients display code

We hold the cl_lock here, and that's enough to keep stateid's from going
away, but it's not enough to prevent the files they point to from going
away.  Take fi_lock and a reference and check for NULL, as we do in
other code.

Reported-by: NeilBrown <neilb@suse.de>
Fixes: 78599c42ae3c ("nfsd4: add file to display list of client's opens")
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
4 years agoMerge tag 'media/v5.8-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Wed, 22 Jul 2020 18:56:00 +0000 (11:56 -0700)]
Merge tag 'media/v5.8-3' of git://git./linux/kernel/git/mchehab/linux-media into master

Pull media fixes from Mauro Carvalho Chehab:
 "A series of fixes for the upcoming atomisp driver. They solve issues
  when probing atomisp on devices with multiple cameras and get rid of
  warnings when built with W=1.

  The diffstat is a bit long, as this driver has several abstractions.
  The patches that solved the issues with W=1 had to get rid of some
  duplicated code (there used to have 2 versions of the same code, one
  for ISP2401 and another one for ISP2400).

  As this driver is not in 5.7, such changes won't cause regressions"

* tag 'media/v5.8-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media: (38 commits)
  Revert "media: atomisp: keep the ISP powered on when setting it"
  media: atomisp: fix mask and shift operation on ISPSSPM0
  media: atomisp: move system_local consts into a C file
  media: atomisp: get rid of version-specific system_local.h
  media: atomisp: move global stuff into a common header
  media: atomisp: remove non-used 32-bits consts at system_local
  media: atomisp: get rid of some unused static vars
  media: atomisp: Fix error code in ov5693_probe()
  media: atomisp: Replace trace_printk by pr_info
  media: atomisp: Fix __func__ style warnings
  media: atomisp: fix help message for ISP2401 selection
  media: atomisp: i2c: atomisp-ov2680.c: fixed a brace coding style issue.
  media: atomisp: make const arrays static, makes object smaller
  media: atomisp: Clean up non-existing folders from Makefile
  media: atomisp: Get rid of ACPI specifics in gmin_subdev_add()
  media: atomisp: Provide Gmin subdev as parameter to gmin_subdev_add()
  media: atomisp: Use temporary variable for device in gmin_subdev_add()
  media: atomisp: Refactor PMIC detection to a separate function
  media: atomisp: Deduplicate return ret in gmin_i2c_write()
  media: atomisp: Make pointer to PMIC client global
  ...

4 years agoMerge tag 'exfat-for-5.8-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/linkin...
Linus Torvalds [Wed, 22 Jul 2020 18:30:07 +0000 (11:30 -0700)]
Merge tag 'exfat-for-5.8-rc7' of git://git./linux/kernel/git/linkinjeon/exfat into master

Pull exfat fixes from Namjae Jeon:

 - fix overflow issue at sector calculation

 - fix wrong hint_stat initialization

 - fix wrong size update of stream entry

 - fix endianness of upname in name_hash computation

* tag 'exfat-for-5.8-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/exfat:
  exfat: fix name_hash computation on big endian systems
  exfat: fix wrong size update of stream entry by typo
  exfat: fix wrong hint_stat initialization in exfat_find_dir_entry()
  exfat: fix overflow issue in exfat_cluster_to_sector()

4 years agostaging: wlan-ng: properly check endpoint types
Rustam Kovhaev [Wed, 22 Jul 2020 16:10:52 +0000 (09:10 -0700)]
staging: wlan-ng: properly check endpoint types

As syzkaller detected, wlan-ng driver does not do sanity check of
endpoints in prism2sta_probe_usb(), add check for xfer direction and type

Reported-and-tested-by: syzbot+c2a1fa67c02faa0de723@syzkaller.appspotmail.com
Link: https://syzkaller.appspot.com/bug?extid=c2a1fa67c02faa0de723
Signed-off-by: Rustam Kovhaev <rkovhaev@gmail.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200722161052.999754-1-rkovhaev@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoRevert "PCI/PM: Assume ports without DLL Link Active train links in 100 ms"
Bjorn Helgaas [Fri, 17 Jul 2020 22:21:28 +0000 (17:21 -0500)]
Revert "PCI/PM: Assume ports without DLL Link Active train links in 100 ms"

This reverts commit ec411e02b7a2e785a4ed9ed283207cd14f48699d.

Patrick reported that this commit broke hybrid graphics on a ThinkPad X1
Extreme 2nd with Intel UHD Graphics 630 and NVIDIA GeForce GTX 1650 Mobile:

  nouveau 0000:01:00.0: fifo: PBDMA0: 01000000 [] ch 0 [00ff992000 DRM] subc 0 mthd 0008 data 00000000

Karol reported that this commit broke Nouveau firmware loading on a Lenovo
P1G2 with Intel UHD Graphics 630 and NVIDIA TU117GLM [Quadro T1000 Mobile]:

  nouveau 0000:01:00.0: acr: AHESASC binary failed

In both cases, reverting ec411e02b7a2 solved the problem.  Unfortunately,
this revert will reintroduce the "Thunderbolt bridges take long time to
resume from D3cold" problem:
https://bugzilla.kernel.org/show_bug.cgi?id=206837

Link: https://lore.kernel.org/r/CAErSpo5sTeK_my1dEhWp7aHD0xOp87+oHYWkTjbL7ALgDbXo-Q@mail.gmail.com
Link: https://lore.kernel.org/r/CACO55tsAEa5GXw5oeJPG=mcn+qxNvspXreJYWDJGZBy5v82JDA@mail.gmail.com
Link: https://bugzilla.kernel.org/show_bug.cgi?id=208597
Reported-by: Patrick Volkerding <volkerdi@gmail.com>
Reported-by: Karol Herbst <kherbst@redhat.com>
Fixes: ec411e02b7a2 ("PCI/PM: Assume ports without DLL Link Active train links in 100 ms")
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
4 years agovirtio-mmio: Reject invalid IRQ 0 command line argument
Bjorn Helgaas [Wed, 1 Jul 2020 20:53:15 +0000 (15:53 -0500)]
virtio-mmio: Reject invalid IRQ 0 command line argument

The "virtio_mmio.device=" command line argument allows a user to specify
the size, address, and IRQ of a virtio device.  Previously the only
requirement for the IRQ was that it be an unsigned integer.

Zero is an unsigned integer but an invalid IRQ number, and after
a85a6c86c25be ("driver core: platform: Clarify that IRQ 0 is invalid"),
attempts to use IRQ 0 cause warnings.

If the user specifies IRQ 0, return failure instead of registering a device
with IRQ 0.

Fixes: a85a6c86c25be ("driver core: platform: Clarify that IRQ 0 is invalid")
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
4 years agoiommu/qcom: Use domain rather than dev as tlb cookie
Rob Clark [Mon, 20 Jul 2020 15:52:17 +0000 (08:52 -0700)]
iommu/qcom: Use domain rather than dev as tlb cookie

The device may be torn down, but the domain should still be valid.  Lets
use that as the tlb flush ops cookie.

Fixes a problem reported in [1]

[1] https://lkml.org/lkml/2020/7/20/104

Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>
Tested-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Fixes: 09b5dfff9ad6 ("iommu/qcom: Use accessor functions for iommu private data")
Link: https://lore.kernel.org/r/20200720155217.274994-1-robdclark@gmail.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
4 years agoMAINTAINERS: add Matthew for s390 IOMMU
Gerald Schaefer [Tue, 21 Jul 2020 13:04:30 +0000 (15:04 +0200)]
MAINTAINERS: add Matthew for s390 IOMMU

Acked-By: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
4 years agousb: usbfs: stop using compat_alloc_user_space
Christoph Hellwig [Wed, 22 Jul 2020 07:36:55 +0000 (09:36 +0200)]
usb: usbfs: stop using compat_alloc_user_space

Just switch the low-level routines to take kernel structures, and do the
conversion from the compat to the native structure on that.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20200722073655.220011-1-hch@lst.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoMAINTAINERS: i2c: at91: handover maintenance to Codrin Ciubotariu
Ludovic Desroches [Thu, 9 Jul 2020 08:42:33 +0000 (10:42 +0200)]
MAINTAINERS: i2c: at91: handover maintenance to Codrin Ciubotariu

My colleague Codrin Ciubotariu, now, maintains this driver internally.
Then I handover the mainline maintenance to him.

Signed-off-by: Ludovic Desroches <ludovic.desroches@microchip.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
4 years agoi2c: drop duplicated word in the header file
Randy Dunlap [Fri, 17 Jul 2020 23:38:15 +0000 (16:38 -0700)]
i2c: drop duplicated word in the header file

Drop the doubled word "be" in a comment.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
4 years agoi2c: cadence: Clear HOLD bit at correct time in Rx path
Raviteja Narayanam [Fri, 3 Jul 2020 13:56:12 +0000 (19:26 +0530)]
i2c: cadence: Clear HOLD bit at correct time in Rx path

There are few issues on Zynq SOC observed in the stress tests causing
timeout errors. Even though all the data is received, timeout error
is thrown. This is due to an IP bug in which the COMP bit in ISR is
not set at end of transfer and completion interrupt is not generated.

This bug is seen on Zynq platforms when the following condition occurs:
Master read & HOLD bit set & Transfer size register reaches '0'.

One workaround is to clear the HOLD bit before the transfer size
register reaches '0'. The current implementation checks for this at
the start of the loop and also only for less than FIFO DEPTH case
(ignoring the equal to case).

So clear the HOLD bit when the data yet to receive is less than or
equal to the FIFO DEPTH. This avoids the IP bug condition.

Signed-off-by: Raviteja Narayanam <raviteja.narayanam@xilinx.com>
Acked-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
4 years agoRevert "i2c: cadence: Fix the hold bit setting"
Raviteja Narayanam [Fri, 3 Jul 2020 13:55:49 +0000 (19:25 +0530)]
Revert "i2c: cadence: Fix the hold bit setting"

This reverts commit d358def706880defa4c9e87381c5bf086a97d5f9.

There are two issues with "i2c: cadence: Fix the hold bit setting" commit.

1. In case of combined message request from user space, when the HOLD
bit is cleared in cdns_i2c_mrecv function, a STOP condition is sent
on the bus even before the last message is started. This is because when
the HOLD bit is cleared, the FIFOS are empty and there is no pending
transfer. The STOP condition should occur only after the last message
is completed.

2. The code added by the commit is redundant. Driver is handling the
setting/clearing of HOLD bit in right way before the commit.

The setting of HOLD bit based on 'bus_hold_flag' is taken care in
cdns_i2c_master_xfer function even before cdns_i2c_msend/cdns_i2c_recv
functions.

The clearing of HOLD bit is taken care at the end of cdns_i2c_msend and
cdns_i2c_recv functions based on bus_hold_flag and byte count.
Since clearing of HOLD bit is done after the slave address is written to
the register (writing to address register triggers the message transfer),
it is ensured that STOP condition occurs at the right time after
completion of the pending transfer (last message).

Signed-off-by: Raviteja Narayanam <raviteja.narayanam@xilinx.com>
Acked-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
4 years agosched: Fix race against ptrace_freeze_trace()
Peter Zijlstra [Mon, 20 Jul 2020 15:20:21 +0000 (17:20 +0200)]
sched: Fix race against ptrace_freeze_trace()

There is apparently one site that violates the rule that only current
and ttwu() will modify task->state, namely ptrace_{,un}freeze_traced()
will change task->state for a remote task.

Oleg explains:

  "TASK_TRACED/TASK_STOPPED was always protected by siglock. In
particular, ttwu(__TASK_TRACED) must be always called with siglock
held. That is why ptrace_freeze_traced() assumes it can safely do
s/TASK_TRACED/__TASK_TRACED/ under spin_lock(siglock)."

This breaks the ordering scheme introduced by commit:

  dbfb089d360b ("sched: Fix loadavg accounting race")

Specifically, the reload not matching no longer implies we don't have
to block.

Simply things by noting that what we need is a LOAD->STORE ordering
and this can be provided by a control dependency.

So replace:

prev_state = prev->state;
raw_spin_lock(&rq->lock);
smp_mb__after_spinlock(); /* SMP-MB */
if (... && prev_state && prev_state == prev->state)
deactivate_task();

with:

prev_state = prev->state;
if (... && prev_state) /* CTRL-DEP */
deactivate_task();

Since that already implies the 'prev->state' load must be complete
before allowing the 'prev->on_rq = 0' store to become visible.

Fixes: dbfb089d360b ("sched: Fix loadavg accounting race")
Reported-by: Jiri Slaby <jirislaby@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Tested-by: Christian Brauner <christian.brauner@ubuntu.com>
4 years agox86, vmlinux.lds: Page-align end of ..page_aligned sections
Joerg Roedel [Tue, 21 Jul 2020 09:34:48 +0000 (11:34 +0200)]
x86, vmlinux.lds: Page-align end of ..page_aligned sections

On x86-32 the idt_table with 256 entries needs only 2048 bytes. It is
page-aligned, but the end of the .bss..page_aligned section is not
guaranteed to be page-aligned.

As a result, objects from other .bss sections may end up on the same 4k
page as the idt_table, and will accidentially get mapped read-only during
boot, causing unexpected page-faults when the kernel writes to them.

This could be worked around by making the objects in the page aligned
sections page sized, but that's wrong.

Explicit sections which store only page aligned objects have an implicit
guarantee that the object is alone in the page in which it is placed. That
works for all objects except the last one. That's inconsistent.

Enforcing page sized objects for these sections would wreckage memory
sanitizers, because the object becomes artificially larger than it should
be and out of bound access becomes legit.

Align the end of the .bss..page_aligned and .data..page_aligned section on
page-size so all objects places in these sections are guaranteed to have
their own page.

[ tglx: Amended changelog ]

Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20200721093448.10417-1-joro@8bytes.org
4 years agonet: ethernet: ti: add NETIF_F_HW_TC hw feature flag for taprio offload
Murali Karicheri [Fri, 17 Jul 2020 12:19:32 +0000 (15:19 +0300)]
net: ethernet: ti: add NETIF_F_HW_TC hw feature flag for taprio offload

Currently drive supports taprio offload which is a tc feature offloaded
to cpsw hardware. So driver has to set the hw feature flag, NETIF_F_HW_TC
in the net device to be compliant. This patch adds the flag.

Fixes: 8127224c2708 ("ethernet: ti: am65-cpsw-qos: add TAPRIO offload support")
Signed-off-by: Murali Karicheri <m-karicheri2@ti.com>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ethernet: ave: Fix error returns in ave_init
Wang Hai [Fri, 17 Jul 2020 02:50:49 +0000 (10:50 +0800)]
net: ethernet: ave: Fix error returns in ave_init

When regmap_update_bits failed in ave_init(), calls of the functions
reset_control_assert() and clk_disable_unprepare() were missed.
Add goto out_reset_assert to do this.

Fixes: 57878f2f4697 ("net: ethernet: ave: add support for phy-mode setting of system controller")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Reviewed-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodrivers/net/wan/x25_asy: Fix to make it work
Xie He [Thu, 16 Jul 2020 23:44:33 +0000 (16:44 -0700)]
drivers/net/wan/x25_asy: Fix to make it work

This driver is not working because of problems of its receiving code.
This patch fixes it to make it work.

When the driver receives an LAPB frame, it should first pass the frame
to the LAPB module to process. After processing, the LAPB module passes
the data (the packet) back to the driver, the driver should then add a
one-byte pseudo header and pass the data to upper layers.

The changes to the "x25_asy_bump" function and the
"x25_asy_data_indication" function are to correctly implement this
procedure.

Also, the "x25_asy_unesc" function ignores any frame that is shorter
than 3 bytes. However the shortest frames are 2-byte long. So we need
to change it to allow 2-byte frames to pass.

Cc: Eric Dumazet <edumazet@google.com>
Cc: Martin Schiller <ms@dev.tdt.de>
Signed-off-by: Xie He <xie.he.0141@gmail.com>
Reviewed-by: Martin Schiller <ms@dev.tdt.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoipvs: fix the connection sync failed in some cases
guodeqing [Thu, 16 Jul 2020 08:12:08 +0000 (16:12 +0800)]
ipvs: fix the connection sync failed in some cases

The sync_thread_backup only checks sk_receive_queue is empty or not,
there is a situation which cannot sync the connection entries when
sk_receive_queue is empty and sk_rmem_alloc is larger than sk_rcvbuf,
the sync packets are dropped in __udp_enqueue_schedule_skb, this is
because the packets in reader_queue is not read, so the rmem is
not reclaimed.

Here I add the check of whether the reader_queue of the udp sock is
empty or not to solve this problem.

Fixes: 2276f58ac589 ("udp: use a separate rx queue for packet reception")
Reported-by: zhouxudong <zhouxudong8@huawei.com>
Signed-off-by: guodeqing <geffrey.guo@huawei.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
4 years agoselftest: txtimestamp: fix net ns entry logic
Paolo Pisati [Tue, 21 Jul 2020 16:17:10 +0000 (18:17 +0200)]
selftest: txtimestamp: fix net ns entry logic

According to 'man 8 ip-netns', if `ip netns identify` returns an empty string,
there's no net namespace associated with current PID: fix the net ns entrance
logic.

Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoxtensa: fix access check in csum_and_copy_from_user
Max Filippov [Tue, 21 Jul 2020 22:00:35 +0000 (15:00 -0700)]
xtensa: fix access check in csum_and_copy_from_user

Commit d341659f470b ("xtensa: switch to providing
csum_and_copy_from_user()") introduced access check, but incorrectly
tested dst instead of src.
Fix access_ok argument in csum_and_copy_from_user.

Cc: Al Viro <viro@zeniv.linux.org.uk>
Fixes: d341659f470b ("xtensa: switch to providing csum_and_copy_from_user()")
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
4 years agoMerge branch 'qed-suppress-irrelevant-error-messages-on-HW-init'
David S. Miller [Tue, 21 Jul 2020 23:07:34 +0000 (16:07 -0700)]
Merge branch 'qed-suppress-irrelevant-error-messages-on-HW-init'

Alexander Lobakin says:

====================
qed: suppress irrelevant error messages on HW init

This raises the verbosity level of several error/warning messages on
driver/module initialization, most of which are false-positives, and
the one actively spamming the log for no reason.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoqed: suppress false-positives interrupt error messages on HW init
Alexander Lobakin [Tue, 21 Jul 2020 14:41:43 +0000 (17:41 +0300)]
qed: suppress false-positives interrupt error messages on HW init

It was found that qed_pglueb_rbc_attn_handler() can produce a lot of
false-positive error detections on driver load/reload (especially after
crashes/recoveries) and spam the kernel log:

[    4.958275] [qed_pglueb_rbc_attn_handler:324()]ICPL error - 00d00ff0
[ 2079.146764] [qed_pglueb_rbc_attn_handler:324()]ICPL error - 00d80ff0
[ 2116.374631] [qed_pglueb_rbc_attn_handler:324()]ICPL error - 00d80ff0
[ 2135.250564] [qed_pglueb_rbc_attn_handler:324()]ICPL error - 00d80ff0
[...]

Reduce the logging level of two false-positive prone error messages from
notice to verbose on initialization (only) to not mix it with real error
attentions while debugging.

Fixes: 666db4862f2d ("qed: Revise load sequence to avoid PCI errors")
Signed-off-by: Alexander Lobakin <alobakin@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoqed: suppress "don't support RoCE & iWARP" flooding on HW init
Alexander Lobakin [Tue, 21 Jul 2020 14:41:42 +0000 (17:41 +0300)]
qed: suppress "don't support RoCE & iWARP" flooding on HW init

Change the verbosity of the "don't support RoCE & iWARP simultaneously"
warning to debug level to stop flooding on driver/hardware initialization:

[    4.783230] qede 01:00.00: Storm FW 8.37.7.0, Management FW 8.52.9.0
[MBI 15.10.6] [eth0]
[    4.810020] [qed_rdma_set_pf_params:2076()]Current day drivers don't
support RoCE & iWARP simultaneously on the same PF. Default to RoCE-only
[    4.861186] qede 01:00.01: Storm FW 8.37.7.0, Management FW 8.52.9.0
[MBI 15.10.6] [eth1]
[    4.893311] [qed_rdma_set_pf_params:2076()]Current day drivers don't
support RoCE & iWARP simultaneously on the same PF. Default to RoCE-only
[    5.181713] qede a1:00.00: Storm FW 8.37.7.0, Management FW 8.52.9.0
[MBI 15.10.6] [eth2]
[    5.224740] [qed_rdma_set_pf_params:2076()]Current day drivers don't
support RoCE & iWARP simultaneously on the same PF. Default to RoCE-only
[    5.276449] qede a1:00.01: Storm FW 8.37.7.0, Management FW 8.52.9.0
[MBI 15.10.6] [eth3]
[    5.318671] [qed_rdma_set_pf_params:2076()]Current day drivers don't
support RoCE & iWARP simultaneously on the same PF. Default to RoCE-only
[    5.369548] qede a1:00.02: Storm FW 8.37.7.0, Management FW 8.52.9.0
[MBI 15.10.6] [eth4]
[    5.411645] [qed_rdma_set_pf_params:2076()]Current day drivers don't
support RoCE & iWARP simultaneously on the same PF. Default to RoCE-only

Fixes: e0a8f9de16fc ("qed: Add iWARP enablement support")
Signed-off-by: Alexander Lobakin <alobakin@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonetdevsim: fix unbalaced locking in nsim_create()
Taehee Yoo [Tue, 21 Jul 2020 14:51:50 +0000 (14:51 +0000)]
netdevsim: fix unbalaced locking in nsim_create()

In the nsim_create(), rtnl_lock() is called before nsim_bpf_init().
If nsim_bpf_init() is failed, rtnl_unlock() should be called,
but it isn't called.
So, unbalanced locking would occur.

Fixes: e05b2d141fef ("netdevsim: move netdev creation/destruction to dev probe")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: microchip: call phy_remove_link_mode during probe
Helmut Grohne [Tue, 21 Jul 2020 11:07:39 +0000 (13:07 +0200)]
net: dsa: microchip: call phy_remove_link_mode during probe

When doing "ip link set dev ... up" for a ksz9477 backed link,
ksz9477_phy_setup is called and it calls phy_remove_link_mode to remove
1000baseT HDX. During phy_remove_link_mode, phy_advertise_supported is
called. Doing so reverts any previous change to advertised link modes
e.g. using a udevd .link file.

phy_remove_link_mode is not meant to be used while opening a link and
should be called during phy probe when the link is not yet available to
userspace.

Therefore move the phy_remove_link_mode calls into
ksz9477_switch_register. It indirectly calls dsa_register_switch, which
creates the relevant struct phy_devices and we update the link modes
right after that. At that time dev->features is already initialized by
ksz9477_switch_detect.

Remove phy_setup from ksz_dev_ops as no users remain.

Link: https://lore.kernel.org/netdev/20200715192722.GD1256692@lunn.ch/
Fixes: 42fc6a4c613019 ("net: dsa: microchip: prepare PHY for proper advertisement")
Signed-off-by: Helmut Grohne <helmut.grohne@intenta.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'hns3-fixes'
David S. Miller [Tue, 21 Jul 2020 22:49:17 +0000 (15:49 -0700)]
Merge branch 'hns3-fixes'

Huazhong Tan says:

====================
net: hns3: fixes for -net

There are some bugfixes for the HNS3 ethernet driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: hns3: fix return value error when query MAC link status fail
Jian Shen [Tue, 21 Jul 2020 11:03:54 +0000 (19:03 +0800)]
net: hns3: fix return value error when query MAC link status fail

Currently, PF queries the MAC link status per second by calling
function hclge_get_mac_link_status(). It return the error code
when failed to send cmdq command to firmware. It's incorrect,
because this return value is used as the MAC link status, which
0 means link down, and none-zero means link up. So fixes it.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Huazhong tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: hns3: fix error handling for desc filling
Yunsheng Lin [Tue, 21 Jul 2020 11:03:53 +0000 (19:03 +0800)]
net: hns3: fix error handling for desc filling

The content of the TX desc is automatically cleared by the HW
when the HW has sent out the packet to the wire. When desc filling
fails in hns3_nic_net_xmit(), it will call hns3_clear_desc() to do
the error handling, which miss zeroing of the TX desc and the
checking if a unmapping is needed.

So add the zeroing and checking in hns3_clear_desc() to avoid the
above problem. Also add DESC_TYPE_UNKNOWN to indicate the info in
desc_cb is not valid, because hns3_nic_reclaim_desc() may treat
the desc_cb->type of zero as packet and add to the sent pkt
statistics accordingly.

Fixes: 76ad4f0ee747 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>