21 Mar, 2019
9 commits
-
Add support for suspend and resume operation for PM in CAAM driver.
When the CAAM goes in suspend, the hardware is considered to do nothing.
On some platforms, the power of the CAAM is not turned off so it keeps
its configuration.On other platforms, it doesn't so it is necessary to save the state of
the CAAM:
- JRs MID
- Address of input and output ringsLimitation:
When the CAAM is powered OFF, it is resetted so the JDKEK and TDKEK
changes. This impacts crypto transforms using MDHA split-keys
which are kept over suspend as they are encrypted with the JDKEK:
- hmac(*) from caamhash.c
- authenc(hmac(*),*) from caamalg.c
- echainiv(authenc(hmac(*),*)) from caamalg.c
The issue was already present in current code so this patch does not
add a regression in this regard.Reviewed-by: Horia Geantă
Signed-off-by: Franck LENORMAND
(cherry picked from commit b90e25f285a65ee8c8433aba7fe8b19b2cdf70b9) -
The structure partid is not suitable to represent the DECO MID register.
This patch replace partid by masterid which is more appropriate.
Reviewed-by: Horia Geantă
Signed-off-by: Franck LENORMAND
(cherry picked from commit 49d6d90809cb04ae3a63e7e87f670014ab5da0a1) -
The previous patch create a DMA issue detected with
DMA debug as the size unmapped is not always the size
mapped.Signed-off-by: Franck LENORMAND
(cherry picked from commit 00bd0d58f4d339d0488c4eb102e34e70edd017ee) -
Roland reports the following issue and provides a root cause analysis:
"On a v4.19 i.MX6 system with IMA and CONFIG_DMA_API_DEBUG enabled, a
warning is generated when accessing files on a filesystem for which IMA
measurement is enabled:------------[ cut here ]------------
WARNING: CPU: 0 PID: 1 at kernel/dma/debug.c:1181 check_for_stack.part.9+0xd0/0x120
caam_jr 2101000.jr0: DMA-API: device driver maps memory from stack [addr=b668049e]
Modules linked in:
CPU: 0 PID: 1 Comm: switch_root Not tainted 4.19.0-20181214-1 #2
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
Backtrace:
[] (dump_backtrace) from [] (show_stack+0x20/0x24)
[] (show_stack) from [] (dump_stack+0xa0/0xcc)
[] (dump_stack) from [] (__warn+0xf0/0x108)
[] (__warn) from [] (warn_slowpath_fmt+0x58/0x74)
[] (warn_slowpath_fmt) from [] (check_for_stack.part.9+0xd0/0x120)
[] (check_for_stack.part.9) from [] (debug_dma_map_page+0x144/0x174)
[] (debug_dma_map_page) from [] (ahash_final_ctx+0x5b4/0xcf0)
[] (ahash_final_ctx) from [] (ahash_final+0x1c/0x20)
[] (ahash_final) from [] (crypto_ahash_op+0x38/0x80)
[] (crypto_ahash_op) from [] (crypto_ahash_final+0x20/0x24)
[] (crypto_ahash_final) from [] (ima_calc_file_hash+0x29c/0xa40)
[] (ima_calc_file_hash) from [] (ima_collect_measurement+0x1dc/0x240)
[] (ima_collect_measurement) from [] (process_measurement+0x4c4/0x6b8)
[] (process_measurement) from [] (ima_file_check+0x88/0xa4)
[] (ima_file_check) from [] (path_openat+0x5d8/0x1364)
[] (path_openat) from [] (do_filp_open+0x84/0xf0)
[] (do_filp_open) from [] (do_open_execat+0x84/0x1b0)
[] (do_open_execat) from [] (__do_execve_file+0x43c/0x890)
[] (__do_execve_file) from [] (sys_execve+0x44/0x4c)
[] (sys_execve) from [] (ret_fast_syscall+0x0/0x28)
---[ end trace 3455789a10e3aefd ]---The cause is that the struct ahash_request *req is created as a
stack-local variable up in the stack (presumably somewhere in the IMA
implementation), then passed down into the CAAM driver, which tries to
dma_single_map the req->result (indirectly via map_seq_out_ptr_result)
in order to make that buffer available for the CAAM to store the result
of the following hash operation.The calling code doesn't know how req will be used by the CAAM driver,
and there could be other such occurrences where stack memory is passed
down to the CAAM driver. Therefore we should rather fix this issue in
the CAAM driver where the requirements are known."Fix this problem by:
-instructing the crypto engine to write the final hash in state->caam_ctx
-subsequently memcpy-ing the final hash into req->resultReported-by: Roland Hieber
Signed-off-by: Horia Geantă
Signed-off-by: Franck LENORMAND
(cherry picked from commit d8e87d0a42ce7ff9d96c4197c8df4b22e181c780) -
If CONFIG_KASAN enabled, gpu kernel driver will throw out the following error:
drivers/mxc/gpu-viv/hal/kernel/gc_hal_kernel_command.c: In function ‘gckCOMMAND_Commit’:
rivers/mxc/gpu-viv/hal/kernel/gc_hal_kernel_command.c:2646:1: error:
the frame size of 2784 bytes is larger than 2048 bytes [-Werror=frame-larger-than=]Signed-off-by: Xianzhong
(cherry picked from commit d5237ed9bca46858875a9b0ef4134ad477832102) -
This reverts commit ae7756f53980f6d6bde1a72ea0798c5341c764fb.
Root cause identified, the issue of stream mode only happens
on imx8mm DDR4 board, because the DDR QoS setting is not correct.
After the Qos setting change to be default value on imx8mm DDR4
board, we can enable stream mode for USB device mode now.Acked-by: Peter Chen
Signed-off-by: Li Jun
(cherry picked from commit 4dad5065f1088197884de992be9adfed228197d9) -
Fixes: 944c01a889d9 ("spi: lpspi: enable runtime pm for lpspi")
Signed-off-by: kbuild test robot
Signed-off-by: Mark Brown -
set skip firmware bss flag when when driver clear firmware buffer.
If driver didn't clear the buffer, and the flag is set.
Something unexpected may happen.
So clear firmware buffer before download firmware every times.Signed-off-by: ming_qian
-
if seq header is not found before user streamoff.
and user streamon again.
the firmware is keeping parsing seq state.
but it don't really parse seq header
because it has been canceled by streamoff.
The firmware needs driver trigger again to start parse next seq.
In this case, driver will send stop cmd to firmware first,
and send start cmd to firmware to restart parse seq header.
And the user need transfer spspps before I frame again.Signed-off-by: ming_qian
20 Mar, 2019
5 commits
-
on i.MX8
We need to add property in DTS node before when we want to use normal
mode on i.MX8 platform. Check in RM, we also can use CBT register not
must CTRL1 register to set bitrate for normal mode.This patch intends to use CBT register to set bitrate for normal mode on
i.MX8. After this, we don't need to modify the DTS node to support
normal mode.Signed-off-by: Joakim Zhang
-
The alias ID must be defined in device tree, because
that will be used as BUS ID to Cortex M4. If the alias ID
not defined, linux kernel will automatically allocate one
ID which might not be the same number used in Cortex M4 and
Cortex M4 will not send msg to I2C controller.So let's add BUG_ON to catch issue as earlier as possible to avoid
wasting efforts.Signed-off-by: Peng Fan
Reviewed-by: Clark Wang -
Remove the delay waiting, add the sync mechanism during partition reset.
In order to avoid the hang at master side during the rpmsg restore
procedure. Re-initialize the first_notify parameter, when rpmsg master
assumes remote processor is dead.
Otherwise, master side maybe hang during the rpmsg restore procedure
in some corner cases.
~14ms is required by M4 to process the MU message from the cold boot.
Set the max wait of MU_SendMessageTimeout to 20ms.Signed-off-by: Richard Zhu
Reviewed-by: Robin Gong
(cherry picked from commit 6b5573a9c645b947688c171b2dbdb6b70b59356c) -
Clear pending interrupt status of partition reboot during rpmsg
driver probe to avoid to unregister/register virtio device again.Tested-by: Clark Wang
Reviewed-by: Richard Zhu
Signed-off-by: Fugang Duan -
in the former patch, the order of allocate dma buffer will be changed.
if there is no enough memory, some error may occur.
so revert the former change, and add lock for alloc and free dmaSigned-off-by: ming_qian
19 Mar, 2019
1 commit
-
Because we add the partition reboot function, and assign all flexcan
pins to M4, A core cannot access flexcan pins for now.LPUART3 uses flexcan pins before, so disable it to avoid this error:
imx8qxp-pinctrl iomuxc: pin_config_set op failed for pin 110Signed-off-by: Clark Wang
18 Mar, 2019
9 commits
-
after seek
replace flush_drv_q() with clear_queue()Signed-off-by: Shijie Qin
-
Add the model name accordingly for imx7ulp EVKB board since
the machine name and/or compatible may be used by user.Reviewed-by: Richard Zhu
Signed-off-by: Fugang Duan -
Add security setting check for socket interface since
stack will check the return value.Signed-off-by: Fugang Duan
Signed-off-by: Marcel Holtmann -
commit 861cb5eb467f ("brcmfmac: Fix access point mode") upstream.
Since commit 1204aa17f3b4 ("brcmfmac: set WIPHY_FLAG_HAVE_AP_SME flag")
the Raspberry Pi 3 A+ (BCM43455) isn't able to operate in AP mode with
hostapd (device_ap_sme=1 use_monitor=0):brcmfmac: brcmf_cfg80211_stop_ap: setting AP mode failed -52
So add the missing mgmt_stypes for AP mode to fix this.
Fixes: 1204aa17f3b4 ("brcmfmac: set WIPHY_FLAG_HAVE_AP_SME flag")
Suggested-by: Arend van Spriel
Signed-off-by: Stefan Wahren
Acked-by: Arend van Spriel
Signed-off-by: Kalle Valo -
Remove the quirk for disable cyw4356 D3 mode since current FW
already support it.
This revert the commit 28db0ac248b3("MLK-20716 PCI: add quirk for
cyw4356 to disable D3 mode")Reviewed-by: Richard Zhu
Signed-off-by: Fugang Duan -
This patch fixes 43455 CRC error while running throughput test with
suspend/resume stress test.The continuous failure messages before system crash:
brcmfmac: brcmf_sdiod_sglist_rw: CMD53 sg block read failed -84
brcmfmac: brcmf_sdio_rxglom: glom read of 25600 bytes failed: -5
brcmfmac: brcmf_sdio_rxfail: abort command, terminate frame
brcmfmac: brcmf_sdiod_sglist_rw: CMD53 sg block read failed -84
brcmfmac: brcmf_sdio_rxglom: glom read of 24576 bytes failed: -5
brcmfmac: brcmf_sdio_rxfail: abort command, terminate frameSigned-off-by: Wright Feng
Signed-off-by: Fugang Duan -
With asynchronous suspend/resume feature, suspend and resume callbacks to
be executed in parallel with each other. It makes bus changes the state to
BRCMF_BUS_DOWN before all brcmf_cfg80211_suspend IOVAR executions.
The same situation also happens in resume procedure and causes PM mode
keeps in PM_MAX after resume. In order to fix the race condition, We add
one second sleep in bus suspend and cfg80211 resume function.Signed-off-by: Wright Feng
Signed-off-by: Fugang Duan -
We got ifp null pointer kernel panic in brcmf_txfinalize after removing
Wi-Fi USB dongle when data was transmitting, The root cause is that
interface was removed before calling brcmf_txfinalize in
brcmf_fws_dequeue_worker and finally caused kernel panic.Signed-off-by: Wright Feng
Signed-off-by: Fugang Duan -
To trunkcate the addtional bytes, if extra bytes are received.
Current code only have a warning and proceed without handling it.
But in one crash dump reported by DVT, these causes the
crash intermittently. So the processing is limit to the skb->len.Signed-off-by: Raveendran Somu
Signed-off-by: Fugang Duan
15 Mar, 2019
11 commits
-
1. extend resolution change function
2. change uFBCInUse to uActiveSeqTag in struct MediaIPFW_Video_SeqInfoSigned-off-by: Shijie Qin
-
1. refine code for set_pic_end_flag and clear_pic_end_flag
2. Reduce 'find buffer NULL' printing level to prevent printing all the time when normal playingSigned-off-by: Shijie Qin
-
It's caused by allocating and freeing dma buffer in the same time.
and there is no any protection for this case.
refine the driver to avoid this case.destroy workqueue before release resource
Signed-off-by: ming_qian
-
the api supplied by videobuf2 is not thread safe.
so in multi thread case, it may cause crash
refine the decoder driver, make sure the api are synchronizedSigned-off-by: ming_qian
-
the api supplied by videobuf2 is not thread safe.
so in multi thread case, it may cause crash
refine the encoder driver, make sure the api are synchronizedSigned-off-by: ming_qian
-
Add dma support for lpspi0 and lpspi2 modules on i.MX8QXP board.
Ensure the lpspi does not use cs-gpio in slave mode.
Signed-off-by: Clark Wang
-
Related issues have been fixed by (e0e542aeb46a MLK-20060-1 spi: lpspi:
fix wrong transmission when don't use CONT). However, the delay between
sending and receiving in slave mode. This causes the value of
FSR_RXCOUNT cannot reflect whether there is still data not sent timely.
So do this judgement by FSR_TXCOUNT.Signed-off-by: Clark Wang
-
In order to prevent stall from happening during the system suspend operations,
the DPU KMS driver should use a freezable and unbound work queue for nonblock
commits to make sure all work items are drained in the freeze phase of the
suspend operations. So, let's hook up a commit function of our own onto
->atomic_commit to replace the original one drm_atomic_helper_commit() which
uses the system_unbound_wq(unfreezable). The new function is almost a copy
of drm_atomic_helper_commit() expect for the work queue replacement.Note that other drivers which use drm_atomic_helper_commit() may have
the same problem during system suspend. The right fix should be at the
DRM atomic core. However, being conservative for now, it is okay to fix the
issue for this driver only.Signed-off-by: Liu Ying
(cherry picked from commit a80ca5f87f047acea57ace9f9365794f99894285) -
In order to prevent stall from happening during the system suspend operations,
DPU KMS would use a freezable and unbound work queue for nonblock commits to
replace the original one in DRM atomic helper, that is, the unfreezable
system_unbound_wq. This patch introduces such a work queue and makes it ready
to be used by the DPU KMS driver.Signed-off-by: Liu Ying
(cherry picked from commit af5a701cf2c169fda987efa29ae882f032fdd084) -
DPU KMS would use a freezable and unbound work queue for nonblock commits
to prevent stall from happening during the system suspend operations in the
coming commits. In order to make sure DCSS KMS has the same nonblock commit
behaviour as before, the two drivers may have a work queue of their own
respectively. So, let's make the work queues be driver specific.Signed-off-by: Liu Ying
(cherry picked from commit 894623c295a647a36dd56b6a76b310970bd9a165) -
We need to set parent clock for some audio clock in dts,
when the system enter suspend, the whole audio subsystem
may be power off, the parent setting should be lost, so
we need to set CLK_SET_PARENT_NOCACHE flag in imx_clk_mux_scu,
to let the operation of reset parent clock in
drivers/soc/imx/pm-domains.c take effect.This issue only occurs on imx8qm, not on imx8qxp, for on imx8qxp
the ADMA subsystem don't enter power off, there is other
device in power on state in suspend.Signed-off-by: Shengjiu Wang
(cherry picked from commit 5e8b48b90effa1aa2b9fbbd8919b131f6ad86946)
14 Mar, 2019
5 commits
-
Change the log message to report "difference is" instead of "error is" to
avoid confusion. This message is just reports the actual pixel clock for
informational purposes. It is not an actual error.Signed-off-by: Oliver Brown
-
mipid_hx8369_get_lcd_videomode and mipid_hx8369_lcd_setup are now exported
for the module build.Signed-off-by: Oliver Brown
-
The reset needs to be set in phy_init to handle the warm reset case.
Signed-off-by: Oliver Brown
-
The resets need to be set in poweron to handle the warm reset case.
Signed-off-by: Oliver Brown
-
For some chips may need long time to get the response from M4 sometimes,
enlarge timeout to 500ms.Add a judgement to check if the received data is the current transfer
wanted.Signed-off-by: Clark Wang
