openwrt/staging/blogic.git
4 years agoiwlwifi: bump FW API to 53 for 22000 series
Luca Coelho [Tue, 1 Oct 2019 06:38:08 +0000 (09:38 +0300)]
iwlwifi: bump FW API to 53 for 22000 series

Start supporting API version 53 for 22000 series.

Change-Id: I5725e46394f3f53c3069723fd513cc53c7df383d
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: remove IWL_FW_DBG_DOMAIN macro
Johannes Berg [Wed, 20 Nov 2019 12:48:15 +0000 (13:48 +0100)]
iwlwifi: remove IWL_FW_DBG_DOMAIN macro

This is used to initialize the default value, but refers back
to the value itself, essentially leading to a
val = val
assignment at init time ... that's useless, remove it and use
_NODEF.

Change-Id: I725923016563c34ce2fa057bf7c12984e1041c49
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: yoyo: enable yoyo by default
Luca Coelho [Thu, 7 Nov 2019 13:59:43 +0000 (15:59 +0200)]
iwlwifi: yoyo: enable yoyo by default

Now that YoYo is more mature, we can enable it by default, so we
collect data in the new way.

Change-Id: Ic1a147f935286b085ca8bdb248a7493b7c6341ea
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: scan: support FW APIs with variable number of profiles
Tova Mussai [Tue, 5 Nov 2019 10:56:14 +0000 (12:56 +0200)]
iwlwifi: scan: support FW APIs with variable number of profiles

The FW changed the maximum number of scan offload profiles to 8 in new
APIs.  Support it by changing the scan_offload_profile_cfg struct to be
more dynamic, so we can reuse most of the code and only change size of
the profiles array.

Change-Id: I65210cf541af46e2675a8c764d5871f7f5b070d6
Signed-off-by: Tova Mussai <tova.mussai@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: yoyo: don't block dumping internal memory when not in SRAM mode
rotem saado [Wed, 13 Nov 2019 12:20:11 +0000 (14:20 +0200)]
iwlwifi: yoyo: don't block dumping internal memory when not in SRAM mode

When we have an assert during D3 we want to dump the internal
buffer memory even if are we working in DRAM debug mode.  We should not
block dumping it.

Change-Id: I69aad2d4904c4f8bb653c61e8781a2e07780054f
Signed-off-by: Rotem Saado <rotem.saado@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: remove support for QnJ HR FPGA
Luca Coelho [Mon, 11 Nov 2019 15:32:08 +0000 (17:32 +0200)]
iwlwifi: remove support for QnJ HR FPGA

We don't support the FPGA versions of this card combination anymore.
Remove the cfg mangling that tries to load it and all the relevant
structures.

Change-Id: I190652101afcab682cfba873d062992f11efca32
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: remove support for QnJ Hr STEP A
Luca Coelho [Mon, 11 Nov 2019 15:22:59 +0000 (17:22 +0200)]
iwlwifi: remove support for QnJ Hr STEP A

We don't support QnJ HR STEP A anymore.  Remove the structures we used
for it.

Change-Id: I0dfd88232bdc8ff2dd9c4368b8ed9a0c40c86bc8
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: add trans_cfg for devices with long latency
Luca Coelho [Mon, 4 Nov 2019 20:51:54 +0000 (22:51 +0200)]
iwlwifi: add trans_cfg for devices with long latency

A couple of SoCs, which can be recognized by PCI device IDs 0xA0F0 and
0x43F0, need a longer wait for the xtal to stabilize.  To handle this,
add a new trans_cfg structure for Qu devices with a larger
xtal_latency value and apply them to the devices recognized by these
IDs.  Also add a flag that allows us to inform the FW that the low
latency xtal should be used.

Change-Id: I8a14c6af45ea14d8e7f1ef38a589158f38d0c0ea
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: add support for version 2 of SOC_CONFIGURATION_CMD
Luca Coelho [Wed, 6 Nov 2019 07:27:51 +0000 (09:27 +0200)]
iwlwifi: add support for version 2 of SOC_CONFIGURATION_CMD

This new command is mostly backwards compatible, with the exception
that the device_type element was changed into a bitmask.  The device
type bit remains the same (because we only had 0 and 1 anyway), but
when using v1 we can't set any other bits, because that would change
the integer.

Other than that, the struct remains the same and the driver can set
the device_type bit in both cases, but it can only set the low_latency
bit if VER_2 is used.

Change-Id: Ib68d4c821ebcce253b42ed0ea15881fb4e3e01da
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: mvm: add support for non EDCA based measurements
Avraham Stern [Sun, 6 Oct 2019 09:28:47 +0000 (12:28 +0300)]
iwlwifi: mvm: add support for non EDCA based measurements

Add support for requesting trigger-based / non trigger-based
measurements.

Change-Id: Ib4d0c471da9c50d9981541a7f5926db384a0f7ce
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: remove unnecessary cfg mangling for Qu C and QuZ with Jf
Luca Coelho [Mon, 4 Nov 2019 19:38:33 +0000 (21:38 +0200)]
iwlwifi: remove unnecessary cfg mangling for Qu C and QuZ with Jf

Now that we identify the correct cfgs with the new tables for Qu step
C and QuZ with Jf, we can remove the mangling we do later on.

Change-Id: Ic01ce67db147e897ad2424f0e05a70a00d2c620e
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: convert QnJ with Jf devices to new config table
Luca Coelho [Mon, 4 Nov 2019 19:29:55 +0000 (21:29 +0200)]
iwlwifi: convert QnJ with Jf devices to new config table

All the QnJ devices have a similar matching to the other Qu devices,
but needs a different configuration.  Convert the QnJ devices to the
new table accordingly.

Change-Id: If236ef3d0da3e605a3379922818f5897e0affd7e
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: convert all Qu with Jf devices to the new config table
Luca Coelho [Mon, 4 Nov 2019 19:02:47 +0000 (21:02 +0200)]
iwlwifi: convert all Qu with Jf devices to the new config table

Add new generic iwl_trans structures for these devices and apply the
correct cfg depending on the device characteristics.

Since we have to match Qu with IWL_CONFIG_ANY, we also need to move
the Hr devices to the new table, but for now we keep matching on PCI
device and subsystem device IDs.

Change-Id: I14e9146a99621ff11ce50bc746a4b88af508fee0
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: add HW step to new cfg device table
Luca Coelho [Mon, 4 Nov 2019 15:29:55 +0000 (17:29 +0200)]
iwlwifi: add HW step to new cfg device table

We need to use different firmware versions for different HW steps with
certain devices.  Prepare for this differentiation by adding HW step
to the new device table.

Change-Id: Ib1afb7b0c89e9dc2d26e6d32ea19e978c17ba1dd
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: remove some unused extern declarations from iwl-config.h
Luca Coelho [Mon, 4 Nov 2019 10:40:13 +0000 (12:40 +0200)]
iwlwifi: remove some unused extern declarations from iwl-config.h

After the refactoring, a few extern declarations were accidentally
left in the iwl-config.h file.  Remove them

Change-Id: I79745e440ed5a0a90db61b0daaae374ecef09e86
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: move integrated, extra_phy and soc_latency to trans_cfg
Luca Coelho [Mon, 4 Nov 2019 10:31:22 +0000 (12:31 +0200)]
iwlwifi: move integrated, extra_phy and soc_latency to trans_cfg

These values are selected based on the PCI device ID, so the decision
to use them can be made early.  By moving them to the trans_cfg, we
avoid duplicating the large cfg structs for small pieces of
data (sometimes a single boolean).  This will also allow us to make
more decisions based on, for instance, the SoC type in used.

The trans_cfg concept changes a bit, because previously it was used
only to boot the device before reading further characteristics and now
it also contains more data that is associated with the device ID.

Change-Id: Ib71b07ea9e322eb74571dc5e8aa58f17eece5c9c
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: mvm: enable SF also when we have HE
Johannes Berg [Thu, 7 Nov 2019 11:04:45 +0000 (12:04 +0100)]
iwlwifi: mvm: enable SF also when we have HE

We shouldn't do this just for HT/VHT, but be future-proof
and also check for HE.

Change-Id: Icaeff714c00a773681dbfee72558afd1c7121c5d
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: remove redundant iwl9560_2ac_cfg struct
Luca Coelho [Mon, 4 Nov 2019 09:53:41 +0000 (11:53 +0200)]
iwlwifi: remove redundant iwl9560_2ac_cfg struct

The iwl9560_2ac_cfg struct is used for PNJ devices and the
configuration is the same as iwl9260_2ac_cfg, so we can remove the
former to avoid redundancy.

Change-Id: I17ac1802f00bd80006930b922a9fc21df60e3c16
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: mvm: rs-fw: fix some indentation
Johannes Berg [Thu, 7 Nov 2019 10:50:55 +0000 (11:50 +0100)]
iwlwifi: mvm: rs-fw: fix some indentation

That closing brace for the switch statement is misplaced, fix it.

Change-Id: I39af135a9e3fc64337d2cced43a70cb48fe3b9c1
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: scan: support scan req cmd ver 14
Shahar S Matityahu [Sun, 29 Sep 2019 09:48:42 +0000 (12:48 +0300)]
iwlwifi: scan: support scan req cmd ver 14

Modify adaptive dwell number of APs override API
Instead of using channel to index mapping, add the adaptive dwell
override parameters as part of the configuration per channel in the scan
request command.

Support 2 different override values and use them as follows:
1. 10 APs for friendly GO channels in p2p scan.
2. 2 APs for social channels in p2p scan.

Change-Id: I3b461108abf2306c3d054099112f2c3afce1cc92
Signed-off-by: Shahar S Matityahu <shahar.s.matityahu@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: pass trans and NVM data to HE capability parsing
Johannes Berg [Tue, 5 Nov 2019 11:22:06 +0000 (12:22 +0100)]
iwlwifi: pass trans and NVM data to HE capability parsing

We'll need this data in the future, pass the values.

Change-Id: Iaeff50716e783f5c0bcea86ca1c93ada1560525e
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: pcie: implement read_config32
Luca Coelho [Fri, 1 Nov 2019 14:13:33 +0000 (16:13 +0200)]
iwlwifi: pcie: implement read_config32

Add the read_config32 op to allow dumping the config space when
needed.

Change-Id: Ib2d254a38a4bfb95dcc3d04eec91781827a0c623
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: yoyo: add PCI config space region type
Luca Coelho [Fri, 1 Nov 2019 13:31:02 +0000 (15:31 +0200)]
iwlwifi: yoyo: add PCI config space region type

Add a new region type that allows us to dump the PCI config space.
This is mostly the same as dumping a memory region, but reading from
the device's config space instead.

In order to make this generic and independent of the trans type, we
make a function called iwl_dump_ini_config_iter() that calls a new op
in the transport to read its config space.

Change-Id: I15151bddf589f13b0e0a45c28b96bbcd73bcfdeb
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agoiwlwifi: mvm: add soc latency support
Shahar S Matityahu [Tue, 20 Sep 2016 14:48:25 +0000 (17:48 +0300)]
iwlwifi: mvm: add soc latency support

Some devices require longer time to stabilize the power and XTAL.
This is especially true for devices integrated in the SoC.  Add
support for a new firmware API that allows the driver to set the
latency value accordingly.

Change-Id: I6829a46b89e4e701f80a0e4033f4dd41ee44ed12
Signed-off-by: Shahar S Matityahu <shahar.s.matityahu@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
4 years agohostap: convert to struct proc_ops
YueHaibing [Thu, 26 Mar 2020 03:24:32 +0000 (11:24 +0800)]
hostap: convert to struct proc_ops

commit 97a32539b956 ("proc: convert everything to "struct proc_ops"")
forget do this convering for prism2_download_aux_dump_proc_fops.

Fixes: 97a32539b956 ("proc: convert everything to "struct proc_ops"")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200326032432.20384-1-yuehaibing@huawei.com
4 years agortw88: fix non-increase management packet sequence number
Tzu-En Huang [Thu, 26 Mar 2020 02:04:08 +0000 (10:04 +0800)]
rtw88: fix non-increase management packet sequence number

In previous setting, management packets' sequence numbers will
not increase and always stay at 0. Add hw sequence number support
for mgmt packets.
The table below shows different sequence number setting in the
tx descriptor.

seq num ctrl      | EN_HWSEQ | DISQSELSEL | HW_SSN_SEL
------------------------------------------------------
sw ctrl           |    0     |    N/A     |    N/A
hw ctrl per MACID |    1     |     0      |    N/A
hw ctrl per HWREG |    1     |     1      |HWREG(0/1/2/3)

Signed-off-by: Tzu-En Huang <tehuang@realtek.com>
Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com>
Reviewed-by: Brian Norris <briannorris@chromium.org>
Tested-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200326020408.25218-1-yhchuang@realtek.com
4 years agoMerge ath-next from git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git
Kalle Valo [Thu, 26 Mar 2020 16:28:37 +0000 (18:28 +0200)]
Merge ath-next from git://git./linux/kernel/git/kvalo/ath.git

ath.git patches for v5.7. Major changes:

ath11k

* handle RX fragments

* enable PN offload

* add support for HE BSS color

4 years agobrcmfmac: add USB autosuspend feature support
Wright Feng [Wed, 25 Mar 2020 08:20:28 +0000 (03:20 -0500)]
brcmfmac: add USB autosuspend feature support

We add enable dynamic suspend (autosuspend) support in host driver, and
it can let platform cut down idle power consumption.
To support autosuspend feature in host driver, kernel need to be built
with CONFIG_USB_SUSPEND and autosuspend need to be turn on.
And we also replace wowl feature with adding "needs_remote_wakeup", so
that host still can be waken by wireless device.

Signed-off-by: Wright Feng <wright.feng@cypress.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1585124429-97371-6-git-send-email-chi-hsien.lin@cypress.com
4 years agobrcmfmac: increase max hanger slots from 1K to 3K in fws layer
Madhan Mohan R [Wed, 25 Mar 2020 08:20:27 +0000 (03:20 -0500)]
brcmfmac: increase max hanger slots from 1K to 3K in fws layer

Will enable FMAC to push more packets to bus tx queue and help
improve throughput when fws queuing is enabled. This change is
required to tune the throughput for passing WMM CERT tests.

Signed-off-by: Madhan Mohan R <madhanmohan.r@cypress.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1585124429-97371-5-git-send-email-chi-hsien.lin@cypress.com
4 years agobrcmfmac: fix the incorrect return value in brcmf_inform_single_bss().
Raveendran Somu [Wed, 25 Mar 2020 08:20:26 +0000 (03:20 -0500)]
brcmfmac: fix the incorrect return value in brcmf_inform_single_bss().

The function brcmf_inform_single_bss returns the value as success,
even when the length exceeds the maximum value.
The fix is to send appropriate code on this error.
This issue is observed when Cypress test group reported random fmac
crashes when running their tests and the path was identified from the
crash logs. With this fix the random failure issue in Cypress test group
was resolved.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Raveendran Somu <raveendran.somu@cypress.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1585124429-97371-4-git-send-email-chi-hsien.lin@cypress.com
4 years agobrcmfmac: Fix double freeing in the fmac usb data path
Raveendran Somu [Wed, 25 Mar 2020 08:20:25 +0000 (03:20 -0500)]
brcmfmac: Fix double freeing in the fmac usb data path

When the brcmf_fws_process_skb() fails to get hanger slot for
queuing the skb, it tries to free the skb.
But the caller brcmf_netdev_start_xmit() of that funciton frees
the packet on error return value.
This causes the double freeing and which caused the kernel crash.

Signed-off-by: Raveendran Somu <raveendran.somu@cypress.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1585124429-97371-3-git-send-email-chi-hsien.lin@cypress.com
4 years agobrcmfmac: Fix driver crash on USB control transfer timeout
Raveendran Somu [Wed, 25 Mar 2020 08:20:24 +0000 (03:20 -0500)]
brcmfmac: Fix driver crash on USB control transfer timeout

When the control transfer gets timed out, the error status
was returned without killing that urb, this leads to using
the same urb. This issue causes the kernel crash as the same
urb is sumbitted multiple times. The fix is to kill the
urb for timeout transfer before returning error

Signed-off-by: Raveendran Somu <raveendran.somu@cypress.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1585124429-97371-2-git-send-email-chi-hsien.lin@cypress.com
4 years agortl8xxxu: Feed current txrate information for mac80211
Chris Chiu [Fri, 20 Mar 2020 06:38:33 +0000 (14:38 +0800)]
rtl8xxxu: Feed current txrate information for mac80211

The nl80211 commands such as 'iw link' can't get current txrate
information from the driver. This commit fills in the tx rate
information from the C2H RA report in the sta_statistics function.

Signed-off-by: Chris Chiu <chiu@endlessm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200320063833.1058-3-chiu@endlessm.com
4 years agortl8xxxu: add enumeration for channel bandwidth
Chris Chiu [Fri, 20 Mar 2020 06:38:32 +0000 (14:38 +0800)]
rtl8xxxu: add enumeration for channel bandwidth

There's a data field in H2C and C2H commands which is used to
carry channel bandwidth information. Add enumeration to make it
more descriptive in code.

Signed-off-by: Chris Chiu <chiu@endlessm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200320063833.1058-2-chiu@endlessm.com
4 years agortw88: add a debugfs entry to enable/disable coex mechanism
Yan-Hsuan Chuang [Fri, 13 Mar 2020 03:30:07 +0000 (11:30 +0800)]
rtw88: add a debugfs entry to enable/disable coex mechanism

Sometimes we need to stop the coex mechanism to debug, so that we
can manually control the device through various outer commands.
Hence, add a new debugfs coex_enable to allow us to enable/disable
the coex mechanism when driver is running.

To disable coex
echo 0 > /sys/kernel/debug/ieee80211/phyX/rtw88/coex_enable

To enable coex
echo 1 > /sys/kernel/debug/ieee80211/phyX/rtw88/coex_enable

To check coex dm is enabled or not
cat /sys/kernel/debug/ieee80211/phyX/rtw88/coex_enable

Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200313033008.20070-3-yhchuang@realtek.com
4 years agortw88: add a debugfs entry to dump coex's info
Yan-Hsuan Chuang [Fri, 13 Mar 2020 03:30:06 +0000 (11:30 +0800)]
rtw88: add a debugfs entry to dump coex's info

Add a new entry "coex_info" in debugfs to dump coex's states for
us to debug on coex's issues.

The basic concept for co-existence (coex, usually for WiFi + BT)
is to decide a strategy based on the current status of WiFi and
BT. So, it means the WiFi driver requires to gather information
from BT side and choose a strategy (TDMA/table/HW settings).

Althrough we can easily check the current status of WiFi, e.g.,
from kernel log or just dump the hardware registers, it is still
very difficult for us to gather so many different types of WiFi
states (such as RFE config, antenna, channel/band, TRX, Power
save). Also we will need BT's information that is stored in
"struct rtw_coex". So it is necessary for us to have a debugfs
that can dump all of the WiFi/BT information required.

Note that to debug on coex related issues, we usually need a
longer period of time of coex_info dump every 2 seconds (for
example, 30 secs, so we should have 15 times of coex_info's
dump).

Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com>
Reviewed-by: Chris Chiu <chiu@endlessm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200313033008.20070-2-yhchuang@realtek.com
4 years agodevlink: expand the devlink-info documentation
Jakub Kicinski [Tue, 24 Mar 2020 17:30:16 +0000 (10:30 -0700)]
devlink: expand the devlink-info documentation

We are having multiple review cycles with all vendors trying
to implement devlink-info. Let's expand the documentation with
more information about what's implemented and motivation behind
this interface in an attempt to make the implementations easier.

Describe what each info section is supposed to contain, and make
some references to other HW interfaces (PCI caps).

Document how firmware management is expected to look, to make
it clear how devlink-info and devlink-flash work in concert.

Name some future work.

v2: - improve wording
v3: - improve wording

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mscc: consolidate a common RGMII delay implementation
Vladimir Oltean [Tue, 24 Mar 2020 14:13:58 +0000 (16:13 +0200)]
net: phy: mscc: consolidate a common RGMII delay implementation

It looks like the VSC8584 PHY driver is rolling its own RGMII delay
configuration code, despite the fact that the logic is mostly the same.

In fact only the register layout and position for the RGMII controls has
changed. So we need to adapt and parameterize the PHY-dependent bit
fields when calling the new generic function.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'axienet-Update-error-handling-and-add-64-bit-DMA-support'
David S. Miller [Tue, 24 Mar 2020 23:33:05 +0000 (16:33 -0700)]
Merge branch 'axienet-Update-error-handling-and-add-64-bit-DMA-support'

Andre Przywara says:

====================
net: axienet: Update error handling and add 64-bit DMA support

a minor update, fixing the 32-bit build breakage, and brightening up
Dave's christmas tree. Rebased against latest net-next/master.

This series is based on net-next as of today (9970de8b013a), which
includes Russell's fixes [1], solving the SGMII issues I have had.

[1] https://lore.kernel.org/netdev/E1j6trA-0003GY-N1@rmk-PC.armlinux.org.uk/

Changelog v2 .. v3:
- Use two "left-shifts by 16" to fix builds with 32-bit phys_addr_t
- reorder variable declarations

Changelog v1 .. v2:
- Add Reviewed-by: tags from Radhey
- Extend kerndoc documentation
- Convert DMA error handler tasklet to work queue
- log DMA mapping errors
- mark DMA mapping error checks as unlikely (in "hot" paths)
- return NETDEV_TX_OK on TX DMA mapping error (increasing TX drop counter)
- Request eth IRQ as an optional IRQ
- Remove no longer needed MDIO IRQ register names
- Drop DT propery check for address width, assume full 64 bit

This series updates the Xilinx Axienet driver to work on our board
here. One big issue was broken SGMII support, which Russell fixed already
(in net-next).
While debugging and understanding the driver, I found several problems
in the error handling and cleanup paths, which patches 2-7 address.
Patch 8 removes a annoying error message, patch 9 paves the way for newer
revisions of the IP. The next patch adds mii-tool support, just for good
measure.

The next four patches add support for 64-bit DMA. This is an integration
option on newer IP revisions (>= v7.1), and expects MSB bits in formerly
reserved registers. Without writing to those MSB registers, the state
machine won't trigger, so it's mandatory to access them, even if they
are zero. Patches 11 and 12 prepare the code by adding accessors, to
wrap this properly and keep it working on older IP revisions.
Patch 13 enables access to the MSB registers, by trying to write a
non-zero value to them and checking if that sticks. Older IP revisions
always read those registers as zero.
Patch 14 then adjusts the DMA mask, based on the autodetected MSB
feature. It uses the full 64 bits in this case, the rest of the system
(actual physical addresses in use) should provide a natural limit if the
chip has connected fewer address lines. If not, the parent DT node can
use a dma-range property.

The Xilinx PG138 and PG021 documents (in versions 7.1 in both cases)
were used for this series.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Allow DMA to beyond 4GB
Andre Przywara [Tue, 24 Mar 2020 13:23:47 +0000 (13:23 +0000)]
net: axienet: Allow DMA to beyond 4GB

With all DMA address accesses wrapped, we can actually support 64-bit
DMA if this option was chosen at IP integration time.
If the IP has been configured for an address width greater than 32 bits,
we assume the full 64 bit DMA width is working. In practise this will be
limited by the actual system address bus width, which will ideally be the
same as the DMA IP address width.
If this is not the case, the actual width can still be configured using a
dma-ranges property in the parent of the MAC node.

This increases the DMA mask on those systems to let the kernel choose
buffers from memory at higher addresses.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Autodetect 64-bit DMA capability
Andre Przywara [Tue, 24 Mar 2020 13:23:46 +0000 (13:23 +0000)]
net: axienet: Autodetect 64-bit DMA capability

When newer revisions of the Axienet IP are configured for a 64-bit bus,
we *need* to write to the MSB part of the an address registers,
otherwise the IP won't recognise this as a DMA start condition.
This is even true when the actual DMA address comes from the lower 4 GB.

To autodetect this configuration, at probe time we write all 1's to such
an MSB register, and see if any bits stick. If this is configured for a
32-bit bus, those MSB registers are RES0, so reading back 0 indicates
that no MSB writes are necessary.
On the other hands reading anything other than 0 indicated the need to
write the MSB registers, so we set the respective flag.

The actual DMA mask stays at 32-bit for now. To help bisecting, a
separate patch will enable allocations from higher addresses.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Upgrade descriptors to hold 64-bit addresses
Andre Przywara [Tue, 24 Mar 2020 13:23:45 +0000 (13:23 +0000)]
net: axienet: Upgrade descriptors to hold 64-bit addresses

Newer revisions of the AXI DMA IP (>= v7.1) support 64-bit addresses,
both for the descriptors itself, as well as for the buffers they are
pointing to.
This is realised by adding "MSB" words for the next and phys pointer
right behind the existing address word, now named "LSB". These MSB words
live in formerly reserved areas of the descriptor.

If the hardware supports it, write both words when setting an address.
The buffer address is handled by two wrapper functions, the two
occasions where we set the next pointers are open coded.

For now this is guarded by a flag which we don't set yet.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Wrap DMA pointer writes to prepare for 64 bit
Andre Przywara [Tue, 24 Mar 2020 13:23:44 +0000 (13:23 +0000)]
net: axienet: Wrap DMA pointer writes to prepare for 64 bit

Newer versions of the Xilink DMA IP support busses with more than 32
address bits, by introducing an MSB word for the registers holding DMA
pointers (tail/current, RX/TX descriptor addresses).
On IP configured for more than 32 bits, it is also *required* to write
both words, to let the IP recognise this as a start condition for an
MM2S request, for instance.

Wrap the DMA pointer writes with a separate function, to add this
functionality later. For now we stick to the lower 32 bits.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Add mii-tool support
Andre Przywara [Tue, 24 Mar 2020 13:23:43 +0000 (13:23 +0000)]
net: axienet: Add mii-tool support

mii-tool is useful for debugging, and all it requires to work is to wire
up the ioctl ops function pointer.
Add this to the axienet driver to enable mii-tool.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Drop MDIO interrupt registers from ethtools dump
Andre Przywara [Tue, 24 Mar 2020 13:23:42 +0000 (13:23 +0000)]
net: axienet: Drop MDIO interrupt registers from ethtools dump

Newer revisions of the IP don't have these registers. Since we don't
really use them, just drop them from the ethtools dump.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Mark eth_irq as optional
Andre Przywara [Tue, 24 Mar 2020 13:23:41 +0000 (13:23 +0000)]
net: axienet: Mark eth_irq as optional

According to the DT binding, the Ethernet core interrupt is optional.

Use platform_get_irq_optional() to avoid the error message when the
IRQ is not specified.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Check for DMA mapping errors
Andre Przywara [Tue, 24 Mar 2020 13:23:40 +0000 (13:23 +0000)]
net: axienet: Check for DMA mapping errors

Especially with the default 32-bit DMA mask, DMA buffers are a limited
resource, so their allocation can fail.
So as the DMA API documentation requires, add error checking code after
dma_map_single() calls to catch the case where we run out of "low" memory.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Factor out TX descriptor chain cleanup
Andre Przywara [Tue, 24 Mar 2020 13:23:39 +0000 (13:23 +0000)]
net: axienet: Factor out TX descriptor chain cleanup

Factor out the code that cleans up a number of connected TX descriptors,
as we will need it to properly roll back a failed _xmit() call.
There are subtle differences between cleaning up a successfully sent
chain (unknown number of involved descriptors, total data size needed)
and a chain that was about to set up (number of descriptors known), so
cater for those variations with some extra parameters.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Improve DMA error handling
Andre Przywara [Tue, 24 Mar 2020 13:23:38 +0000 (13:23 +0000)]
net: axienet: Improve DMA error handling

Since 0 is a valid DMA address, we cannot use the physical address to
check whether a TX descriptor is valid and is holding a DMA mapping.

Use the "cntrl" member of the descriptor to make this decision, as it
contains at least the length of the buffer, so 0 points to an
uninitialised buffer.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Fix DMA descriptor cleanup path
Andre Przywara [Tue, 24 Mar 2020 13:23:37 +0000 (13:23 +0000)]
net: axienet: Fix DMA descriptor cleanup path

When axienet_dma_bd_init() bails out during the initialisation process,
it might do so with parts of the structure already allocated and
initialised, while other parts have not been touched yet. Before
returning in this case, we call axienet_dma_bd_release(), which does not
take care of this corner case.
This is most obvious by the first loop happily dereferencing
lp->rx_bd_v, which we actually check to be non NULL *afterwards*.

Make sure we only unmap or free already allocated structures, by:
- directly returning with -ENOMEM if nothing has been allocated at all
- checking for lp->rx_bd_v to be non-NULL *before* using it
- only unmapping allocated DMA RX regions

This avoids NULL pointer dereferences when initialisation fails.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Propagate failure of DMA descriptor setup
Andre Przywara [Tue, 24 Mar 2020 13:23:36 +0000 (13:23 +0000)]
net: axienet: Propagate failure of DMA descriptor setup

When we fail allocating the DMA buffers in axienet_dma_bd_init(), we
report this error, but carry on with initialisation nevertheless.

This leads to a kernel panic when the driver later wants to send a
packet, as it uses uninitialised data structures.

Make the axienet_device_reset() routine return an error value, as it
contains the DMA buffer initialisation. Make sure we propagate the error
up the chain and eventually fail the driver initialisation, to avoid
relying on non-initialised buffers.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: axienet: Convert DMA error handler to a work queue
Andre Przywara [Tue, 24 Mar 2020 13:23:35 +0000 (13:23 +0000)]
net: axienet: Convert DMA error handler to a work queue

The DMA error handler routine is currently a tasklet, scheduled to run
after the DMA error IRQ was handled.
However it needs to take the MDIO mutex, which is not allowed to do in a
tasklet. A kernel (with debug options) complains consequently:
[  614.050361] net eth0: DMA Tx error 0x174019
[  614.064002] net eth0: Current BD is at: 0x8f84aa0ce
[  614.080195] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:935
[  614.109484] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 40, name: kworker/u4:4
[  614.135428] 3 locks held by kworker/u4:4/40:
[  614.149075]  #0: ffff000879863328 ((wq_completion)rpciod){....}, at: process_one_work+0x1f0/0x6a8
[  614.177528]  #1: ffff80001251bdf8 ((work_completion)(&task->u.tk_work)){....}, at: process_one_work+0x1f0/0x6a8
[  614.209033]  #2: ffff0008784e0110 (sk_lock-AF_INET-RPC){....}, at: tcp_sendmsg+0x24/0x58
[  614.235429] CPU: 0 PID: 40 Comm: kworker/u4:4 Not tainted 5.6.0-rc3-00926-g4a165a9d5921 #26
[  614.260854] Hardware name: ARM Test FPGA (DT)
[  614.274734] Workqueue: rpciod rpc_async_schedule
[  614.289022] Call trace:
[  614.296871]  dump_backtrace+0x0/0x1a0
[  614.308311]  show_stack+0x14/0x20
[  614.318751]  dump_stack+0xbc/0x100
[  614.329403]  ___might_sleep+0xf0/0x140
[  614.341018]  __might_sleep+0x4c/0x80
[  614.352201]  __mutex_lock+0x5c/0x8a8
[  614.363348]  mutex_lock_nested+0x1c/0x28
[  614.375654]  axienet_dma_err_handler+0x38/0x388
[  614.389999]  tasklet_action_common.isra.15+0x160/0x1a8
[  614.405894]  tasklet_action+0x24/0x30
[  614.417297]  efi_header_end+0xe0/0x494
[  614.429020]  irq_exit+0xd0/0xd8
[  614.439047]  __handle_domain_irq+0x60/0xb0
[  614.451877]  gic_handle_irq+0xdc/0x2d0
[  614.463486]  el1_irq+0xcc/0x180
[  614.473451]  __tcp_transmit_skb+0x41c/0xb58
[  614.486513]  tcp_write_xmit+0x224/0x10a0
[  614.498792]  __tcp_push_pending_frames+0x38/0xc8
[  614.513126]  tcp_rcv_established+0x41c/0x820
[  614.526301]  tcp_v4_do_rcv+0x8c/0x218
[  614.537784]  __release_sock+0x5c/0x108
[  614.549466]  release_sock+0x34/0xa0
[  614.560318]  tcp_sendmsg+0x40/0x58
[  614.571053]  inet_sendmsg+0x40/0x68
[  614.582061]  sock_sendmsg+0x18/0x30
[  614.593074]  xs_sendpages+0x218/0x328
[  614.604506]  xs_tcp_send_request+0xa0/0x1b8
[  614.617461]  xprt_transmit+0xc8/0x4f0
[  614.628943]  call_transmit+0x8c/0xa0
[  614.640028]  __rpc_execute+0xbc/0x6f8
[  614.651380]  rpc_async_schedule+0x28/0x48
[  614.663846]  process_one_work+0x298/0x6a8
[  614.676299]  worker_thread+0x40/0x490
[  614.687687]  kthread+0x134/0x138
[  614.697804]  ret_from_fork+0x10/0x18
[  614.717319] xilinx_axienet 7fe00000.ethernet eth0: Link is Down
[  615.748343] xilinx_axienet 7fe00000.ethernet eth0: Link is Up - 1Gbps/Full - flow control off

Since tasklets are not really popular anymore anyway, lets convert this
over to a work queue, which can sleep and thus can take the MDIO mutex.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: xilinx: temac: Relax Kconfig dependencies
Andre Przywara [Tue, 24 Mar 2020 13:23:34 +0000 (13:23 +0000)]
net: xilinx: temac: Relax Kconfig dependencies

Similar to axienet, the temac driver is now architecture agnostic, and
can be at least compiled for several architectures.
Especially the fact that this is a soft IP for implementing in FPGAs
makes the current restriction rather pointless, as it could literally
appear on any architecture, as long as an FPGA is connected to the bus.

The driver hasn't been actually tried on any hardware, it is just a
drive-by patch when doing the same for axienet (a similar patch for
axienet is already merged).

This (temac and axienet) have been compile-tested for:
alpha hppa64 microblaze mips64 powerpc powerpc64 riscv64 s390 sparc64
(using kernel.org cross compilers).

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoethtool: fix incorrect tx-checksumming settings reporting
Vladyslav Tarasiuk [Tue, 24 Mar 2020 11:57:08 +0000 (13:57 +0200)]
ethtool: fix incorrect tx-checksumming settings reporting

Currently, ethtool feature mask for checksum command is ORed with
NETIF_F_FCOE_CRC_BIT, which is bit's position number, instead of the
actual feature bit - NETIF_F_FCOE_CRC.

The invalid bitmask here might affect unrelated features when toggling
TX checksumming. For example, TX checksumming is always mistakenly
reported as enabled on the netdevs tested (mlx5, virtio_net).

Fixes: f70bb06563ed ("ethtool: update mapping of features to legacy ioctl requests")
Signed-off-by: Vladyslav Tarasiuk <vladyslavt@mellanox.com>
Reviewed-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mdio-mux-bcm-iproc: use readl_poll_timeout() to simplify code
Dejin Zheng [Tue, 24 Mar 2020 11:26:47 +0000 (19:26 +0800)]
net: phy: mdio-mux-bcm-iproc: use readl_poll_timeout() to simplify code

use readl_poll_timeout() to replace the poll codes for simplify
iproc_mdio_wait_for_idle() function

Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'wireless-drivers-next-2020-03-24' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Tue, 24 Mar 2020 23:15:58 +0000 (16:15 -0700)]
Merge tag 'wireless-drivers-next-2020-03-24' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for v5.7

Second set of patches for v5.7. Lots of cleanup patches this time, but
of course various new features as well fixes.

When merging with wireless-drivers this pull request has a conflict in:

drivers/net/wireless/intel/iwlwifi/pcie/drv.c

To solve that just drop the changes from commit cf52c8a776d1 in
wireless-drivers and take the hunk from wireless-drivers-next as is.
The list of specific subsystem device IDs are not necessary after
commit d6f2134a3831 (in wireless-drivers-next) anymore, the detection
is based on other characteristics of the devices.

Major changes:

qtnfmac

* support WPA3 SAE and OWE in AP mode

ath10k

* support for getting btcoex settings from Device Tree

* support QCA9377 SDIO device

ath11k

* add HE rate accounting

* add thermal sensor and cooling devices

mt76

* MT7663 support for the MT7615 driver
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'PTP_CLK-pin-configuration-for-SJA1105-DSA-driver'
David S. Miller [Tue, 24 Mar 2020 05:15:07 +0000 (22:15 -0700)]
Merge branch 'PTP_CLK-pin-configuration-for-SJA1105-DSA-driver'

Vladimir Oltean says:

====================
PTP_CLK pin configuration for SJA1105 DSA driver

This series adds support for the PTP_CLK pin on SJA1105 to be configured
via the PTP subsystem, in the "periodic output" and "external timestamp
input" modes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: sja1105: configure the PTP_CLK pin as EXT_TS or PER_OUT
Vladimir Oltean [Mon, 23 Mar 2020 22:59:24 +0000 (00:59 +0200)]
net: dsa: sja1105: configure the PTP_CLK pin as EXT_TS or PER_OUT

The SJA1105 switch family has a PTP_CLK pin which emits a signal with
fixed 50% duty cycle, but variable frequency and programmable start time.

On the second generation (P/Q/R/S) switches, this pin supports even more
functionality. The use case described by the hardware documents talks
about synchronization via oneshot pulses: given 2 sja1105 switches,
arbitrarily designated as a master and a slave, the master emits a
single pulse on PTP_CLK, while the slave is configured to timestamp this
pulse received on its PTP_CLK pin (which must obviously be configured as
input). The difference between the timestamps then exactly becomes the
slave offset to the master.

The only trouble with the above is that the hardware is very much tied
into this use case only, and not very generic beyond that:
 - When emitting a oneshot pulse, instead of being told when to emit it,
   the switch just does it "now" and tells you later what time it was,
   via the PTPSYNCTS register. [ Incidentally, this is the same register
   that the slave uses to collect the ext_ts timestamp from, too. ]
 - On the sync slave, there is no interrupt mechanism on reception of a
   new extts, and no FIFO to buffer them, because in the foreseen use
   case, software is in control of both the master and the slave pins,
   so it "knows" when there's something to collect.

These 2 problems mean that:
 - We don't support (at least yet) the quirky oneshot mode exposed by
   the hardware, just normal periodic output.
 - We abuse the hardware a little bit when we expose generic extts.
   Because there's no interrupt mechanism, we need to poll at double the
   frequency we expect to receive a pulse. Currently that means a
   non-configurable "twice a second".

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: sja1105: make the AVB table dynamically reconfigurable
Vladimir Oltean [Mon, 23 Mar 2020 22:59:23 +0000 (00:59 +0200)]
net: dsa: sja1105: make the AVB table dynamically reconfigurable

The AVB table contains the CAS_MASTER field (to be added in the next
patch) which decides the direction of the PTP_CLK pin.

Reconfiguring this field dynamically is highly preferable to having to
reset the switch and upload a new static configuration, so we add
support for exactly that.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: sja1105: make future_base_time a common helper
Vladimir Oltean [Mon, 23 Mar 2020 22:59:22 +0000 (00:59 +0200)]
net: dsa: sja1105: make future_base_time a common helper

Because the PTP_CLK pin starts toggling only at a time higher than the
current PTP clock, this helper from the time-aware shaper code comes in
handy here as well. We'll use it to transform generic user input for the
perout request into valid input for the sja1105 hardware.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: sja1105: unconditionally set DESTMETA and SRCMETA in AVB table
Vladimir Oltean [Mon, 23 Mar 2020 22:59:21 +0000 (00:59 +0200)]
net: dsa: sja1105: unconditionally set DESTMETA and SRCMETA in AVB table

These fields configure the destination and source MAC address that the
switch will put in the Ethernet frames sent towards the CPU port that
contain RX timestamps for PTP.

These fields do not enable the feature itself, that is configured via
SEND_META0 and SEND_META1 in the General Params table.

The implication of this patch is that the AVB Params table will always
be present in the static config. Which doesn't really hurt.

This is needed because in a future patch, we will add another field from
this table, CAS_MASTER, for configuring the PTP_CLK pin function. That
can be configured irrespective of whether RX timestamping is enabled or
not, so always having this table present is going to simplify things a
bit.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: typhoon: Add required whitespace after keywords
Logan Magee [Mon, 23 Mar 2020 21:31:10 +0000 (13:31 -0800)]
net: typhoon: Add required whitespace after keywords

checkpatch found a lack of appropriate whitespace after certain keywords
as per the style guide. Add it in.

Signed-off-by: Logan Magee <mageelog@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'introduce-read_poll_timeout'
David S. Miller [Tue, 24 Mar 2020 05:00:03 +0000 (22:00 -0700)]
Merge branch 'introduce-read_poll_timeout'

Dejin Zheng says:

====================
introduce read_poll_timeout

This patch sets is introduce read_poll_timeout macro, it is an extension
of readx_poll_timeout macro. the accessor function op just supports only
one parameter in the readx_poll_timeout macro, but this macro can
supports multiple variable parameters for it. so functions like
phy_read(struct phy_device *phydev, u32 regnum) and
phy_read_mmd(struct phy_device *phydev, int devad, u32 regnum) can
use this poll timeout framework.

the first patch introduce read_poll_timeout macro, and the second patch
redefined readx_poll_timeout macro by read_poll_timeout(), and the other
patches are examples using read_poll_timeout macro.

v6 -> v7:
- add a parameter to supports that it can sleep some time
  before read operation in read_poll_timeout macro.
- add prefix with double underscores for some variable to avoid
  any variable re-declaration or shadowing in patch 3 and patch
  7.
v5 -> v6:
- add some check to keep the code more similar in patch 8
v4 -> v5:
- add some msleep() before call phy_read_mmd_poll_timeout() to
  keep the code more similar in patch 6 and patch 9.
- add a patch of drop by v4, it can add msleep before call
  phy_read_poll_timeout() to keep the code more similar.
v3 -> v4:
- add 3 examples of using new functions.
- deal with precedence issues for parameter cond.
- drop a patch about phy_poll_reset() function.
v2 -> v3:
- modify the parameter order of newly added functions.
  phy_read_mmd_poll_timeout(val, cond, sleep_us, timeout_us, \
     phydev, devaddr, regnum)
||
\/
  phy_read_mmd_poll_timeout(phydev, devaddr regnum, val, cond, \
    sleep_us, timeout_us)

  phy_read_poll_timeout(val, cond, sleep_us, timeout_us, \
phydev, regnum)
||
\/
  phy_read_poll_timeout(phydev, regnum, val, cond, sleep_us, \
timeout_us)
v1 -> v2:
- passed a phydev, device address and a reg to replace args...
  parameter in phy_read_mmd_poll_timeout() by Andrew Lunn 's
  suggestion in patch 3. Andrew Lunn <andrew@lunn.ch>, Thanks
  very much for your help!
- also in patch 3, handle phy_read_mmd return an error(the return
  value < 0) in phy_read_mmd_poll_timeout(). Thanks Andrew
  again.
- in patch 6, pass a phydev and a reg to replace args...
  parameter in phy_read_poll_timeout(), and also handle the
  phy_read() function's return error.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: tja11xx: use phy_read_poll_timeout() to simplify the code
Dejin Zheng [Mon, 23 Mar 2020 15:06:00 +0000 (23:06 +0800)]
net: phy: tja11xx: use phy_read_poll_timeout() to simplify the code

use phy_read_poll_timeout() to replace the poll codes for
simplify tja11xx_check() function.

Suggested-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: smsc: use phy_read_poll_timeout() to simplify the code
Dejin Zheng [Mon, 23 Mar 2020 15:05:59 +0000 (23:05 +0800)]
net: phy: smsc: use phy_read_poll_timeout() to simplify the code

use phy_read_poll_timeout() to replace the poll codes for
simplify lan87xx_read_status() function.

Suggested-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: use phy_read_poll_timeout() to simplify the code
Dejin Zheng [Mon, 23 Mar 2020 15:05:58 +0000 (23:05 +0800)]
net: phy: use phy_read_poll_timeout() to simplify the code

use phy_read_poll_timeout() to replace the poll codes for
simplify the code in phy_poll_reset() function.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: introduce phy_read_poll_timeout macro
Dejin Zheng [Mon, 23 Mar 2020 15:05:57 +0000 (23:05 +0800)]
net: phy: introduce phy_read_poll_timeout macro

it is sometimes necessary to poll a phy register by phy_read()
function until its value satisfies some condition. introduce
phy_read_poll_timeout() macros that do this.

Suggested-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: marvell10g: use phy_read_mmd_poll_timeout() to simplify the code
Dejin Zheng [Mon, 23 Mar 2020 15:05:56 +0000 (23:05 +0800)]
net: phy: marvell10g: use phy_read_mmd_poll_timeout() to simplify the code

use phy_read_mmd_poll_timeout() to replace the poll codes for
simplify mv3310_reset() function.

Suggested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: aquantia: use phy_read_mmd_poll_timeout() to simplify the code
Dejin Zheng [Mon, 23 Mar 2020 15:05:55 +0000 (23:05 +0800)]
net: phy: aquantia: use phy_read_mmd_poll_timeout() to simplify the code

use phy_read_mmd_poll_timeout() to replace the poll codes for
simplify aqr107_wait_reset_complete() function.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: bcm84881: use phy_read_mmd_poll_timeout() to simplify the code
Dejin Zheng [Mon, 23 Mar 2020 15:05:54 +0000 (23:05 +0800)]
net: phy: bcm84881: use phy_read_mmd_poll_timeout() to simplify the code

use phy_read_mmd_poll_timeout() to replace the poll codes for
simplify bcm84881_wait_init() function.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: introduce phy_read_mmd_poll_timeout macro
Dejin Zheng [Mon, 23 Mar 2020 15:05:53 +0000 (23:05 +0800)]
net: phy: introduce phy_read_mmd_poll_timeout macro

it is sometimes necessary to poll a phy register by phy_read_mmd()
function until its value satisfies some condition. introduce
phy_read_mmd_poll_timeout() macros that do this.

Suggested-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoiopoll: redefined readx_poll_timeout macro to simplify the code
Dejin Zheng [Mon, 23 Mar 2020 15:05:52 +0000 (23:05 +0800)]
iopoll: redefined readx_poll_timeout macro to simplify the code

redefined readx_poll_timeout macro by read_poll_timeout to
simplify the code.

Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoiopoll: introduce read_poll_timeout macro
Dejin Zheng [Mon, 23 Mar 2020 15:05:51 +0000 (23:05 +0800)]
iopoll: introduce read_poll_timeout macro

this macro is an extension of readx_poll_timeout macro. the accessor
function op just supports only one parameter in the readx_poll_timeout
macro, but this macro can supports multiple variable parameters for
it. so functions like phy_read(struct phy_device *phydev, u32 regnum)
and phy_read_mmd(struct phy_device *phydev, int devad, u32 regnum) can
also use this poll timeout core. and also expand it can sleep some time
before read operation.

Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoRemove DST_HOST
David Laight [Mon, 23 Mar 2020 14:31:19 +0000 (14:31 +0000)]
Remove DST_HOST

Previous changes to the IP routing code have removed all the
tests for the DS_HOST route flag.
Remove the flags and all the code that sets it.

Signed-off-by: David Laight <david.laight@aculab.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: thunderx: remove set but not used variable 'tail'
Zheng zengkai [Mon, 23 Mar 2020 06:51:16 +0000 (14:51 +0800)]
net: thunderx: remove set but not used variable 'tail'

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/cavium/thunder/nicvf_queues.c: In function nicvf_sq_free_used_descs:
drivers/net/ethernet/cavium/thunder/nicvf_queues.c:1182:12: warning:
 variable tail set but not used [-Wunused-but-set-variable]

It's not used since commit 4863dea3fab01("net: Adding support for Cavium ThunderX network controller"),
so remove it.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zheng zengkai <zhengzengkai@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: Implement flow dissection for tag_brcm.c
Florian Fainelli [Sun, 22 Mar 2020 21:09:57 +0000 (14:09 -0700)]
net: dsa: Implement flow dissection for tag_brcm.c

Provide a flow_dissect callback which returns the network offset and
where to find the skb protocol, given the tags structure a common
function works for both tagging formats that are supported.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'devlink-Preparations-for-trap-policers-support'
David S. Miller [Tue, 24 Mar 2020 04:40:40 +0000 (21:40 -0700)]
Merge branch 'devlink-Preparations-for-trap-policers-support'

Ido Schimmel says:

====================
devlink: Preparations for trap policers support

This patch set prepares the code for devlink-trap policer support in a
follow-up patch set [1][2]. No functional changes intended.

Policers are going to be added as attributes of packet trap groups,
which are entities used to aggregate logically related packet traps.
This will allow users, for example, to limit all the packets that
encountered an exception during routing to 10Kpps.

However, currently, device drivers register their packet trap groups
implicitly when they register their packet traps via
devlink_traps_register(). This makes it difficult to pass additional
attributes for the groups. For example, the policer bound to the group.

Therefore, this patch set converts device drivers to explicitly register
their packet trap groups. This will later allow these drivers to
register the group with additional attributes, if any.
====================

Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Only pass packet trap group identifier in trap structure
Ido Schimmel [Sun, 22 Mar 2020 18:48:30 +0000 (20:48 +0200)]
devlink: Only pass packet trap group identifier in trap structure

Packet trap groups are now explicitly registered by drivers and not
implicitly registered when the packet traps are registered. Therefore,
there is no need to encode entire group structure the trap is associated
with inside the trap structure.

Instead, only pass the group identifier. Refer to it as initial group
identifier, as future patches will allow user space to move traps
between groups.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Stop reference counting packet trap groups
Ido Schimmel [Sun, 22 Mar 2020 18:48:29 +0000 (20:48 +0200)]
devlink: Stop reference counting packet trap groups

Now that drivers explicitly register their supported packet trap groups
there is no for devlink to create them on-demand and destroy them when
their reference count reaches zero.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonetdevsim: Explicitly register packet trap groups
Ido Schimmel [Sun, 22 Mar 2020 18:48:28 +0000 (20:48 +0200)]
netdevsim: Explicitly register packet trap groups

Use the previously added API to explicitly register / unregister
supported packet trap groups. This is in preparation for future patches
that will enable drivers to pass additional group attributes, such as
associated policer identifier.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_trap: Explicitly register packet trap groups
Ido Schimmel [Sun, 22 Mar 2020 18:48:27 +0000 (20:48 +0200)]
mlxsw: spectrum_trap: Explicitly register packet trap groups

Use the previously added API to explicitly register / unregister
supported packet trap groups. This is in preparation for future patches
that will enable drivers to pass additional group attributes, such as
associated policer identifier.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Add API to register packet trap groups
Ido Schimmel [Sun, 22 Mar 2020 18:48:26 +0000 (20:48 +0200)]
devlink: Add API to register packet trap groups

Currently, packet trap groups are implicitly registered by drivers upon
packet trap registration. When the traps are registered, each is
associated with a group and the group is created by devlink, if it does
not exist already.

This makes it difficult for drivers to pass additional attributes for
the groups.

Therefore, as a preparation for future patches that require passing
additional group attributes, add an API to explicitly register /
unregister these groups.

Next patches will convert existing drivers to use this API.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'r8169-improvements-for-scheduled-task-handling'
David S. Miller [Tue, 24 Mar 2020 04:38:21 +0000 (21:38 -0700)]
Merge branch 'r8169-improvements-for-scheduled-task-handling'

Heiner Kallweit says:

====================
r8169: improvements for scheduled task handling

This series includes some improvements for handling of scheduled tasks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agor8169: improve RTL8168b FIFO overflow workaround
Heiner Kallweit [Sun, 22 Mar 2020 18:03:56 +0000 (19:03 +0100)]
r8169: improve RTL8168b FIFO overflow workaround

So far only the reset bit it set, but the handler executing the reset
is not scheduled. Therefore nothing will happen until some other action
schedules the handler. Improve this by ensuring that the handler is
scheduled.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agor8169: improve rtl_schedule_task
Heiner Kallweit [Sun, 22 Mar 2020 18:03:06 +0000 (19:03 +0100)]
r8169: improve rtl_schedule_task

The current implementation makes the implicit assumption that if a bit
is set, then the work is scheduled already. Remove the need for this
implicit assumption and call schedule_work() always. It will check
internally whether the work is scheduled already.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agor8169: simplify rtl_task
Heiner Kallweit [Sun, 22 Mar 2020 18:02:32 +0000 (19:02 +0100)]
r8169: simplify rtl_task

Currently rtl_task() is designed to handle a large number of tasks.
However we have just one, so we can remove some overhead.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agor8169: add new helper rtl8168g_enable_gphy_10m
Heiner Kallweit [Sat, 21 Mar 2020 18:08:09 +0000 (19:08 +0100)]
r8169: add new helper rtl8168g_enable_gphy_10m

Factor out setting GPHY 10M to new helper rtl8168g_enable_gphy_10m.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Tue, 24 Mar 2020 04:21:33 +0000 (21:21 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
100GbE Intel Wired LAN Driver Updates 2020-03-21

Implement basic support for the devlink interface in the ice driver.
Additionally pave some necessary changes for adding a devlink region that
exposes the NVM contents.

This series first contains 5 patches for enabling and implementing full NVM
read access via the ETHTOOL_GEEPROM interface. This includes some cleanup of
endian-types, a new function for reading from the NVM and Shadow RAM as a flat
addressable space, a function to calculate the available flash size during
load, and a change to how some of the NVM version fields are stored in the
ice_nvm_info structure.

Following this is 3 patches for implementing devlink support. First, one patch
which implements the basic framework and introduces the ice_devlink.c file.
Second, a patch to implement basic .info_get support. Finally, a patch which
reads the device PBA identifier and reports it as the `board.id` value in the
.info_get response.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'octeontx2-vf-Add-network-driver-for-virtual-function'
David S. Miller [Tue, 24 Mar 2020 04:11:44 +0000 (21:11 -0700)]
Merge branch 'octeontx2-vf-Add-network-driver-for-virtual-function'

Sunil Goutham says:

====================
octeontx2-vf: Add network driver for virtual function

This patch series adds  network driver for the virtual functions of
OcteonTX2 SOC's resource virtualization unit (RVU).

Changes from v3:
   * Removed missed out EXPORT symbols in VF driver.

Changes from v2:
   * Removed Copyright license text.
   * Removed wrapper fn()s around mutex_lock and unlock.
   * Got rid of using macro with 'return'.
   * Removed __weak fn()s.
        - Sugested by Leon Romanovsky and Andrew Lunn

Changes from v1:
   * Removed driver version and fixed authorship
   * Removed driver version and fixed authorship in the already
     upstreamed AF, PF drivers.
   * Removed unnecessary checks in sriov_enable and xmit fn()s.
   * Removed WQ_MEM_RECLAIM flag while creating workqueue.
   * Added lock in tx_timeout task.
   * Added 'supported_coalesce_params' in ethtool ops.
   * Minor other cleanups.
        - Sugested by Jakub Kicinski
====================

Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-pf: Remove wrapper APIs for mutex lock and unlock
Sunil Goutham [Fri, 20 Mar 2020 18:57:26 +0000 (00:27 +0530)]
octeontx2-pf: Remove wrapper APIs for mutex lock and unlock

This patch removes wrapper fn()s around mutex_init/lock/unlock.

Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-af: Remove driver version and fix authorship
Sunil Goutham [Fri, 20 Mar 2020 18:57:25 +0000 (00:27 +0530)]
octeontx2-af: Remove driver version and fix authorship

Removed MODULE_VERSION and fixed MODULE_AUTHOR.

Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-pf: Cleanup all receive buffers in SG descriptor
Geetha sowjanya [Fri, 20 Mar 2020 18:57:24 +0000 (00:27 +0530)]
octeontx2-pf: Cleanup all receive buffers in SG descriptor

With MTU sized receive buffers it is not expected to have CQE_RX
with multiple receive buffer pointers. But since same physcial link
is shared by PF and it's VFs, the max receive packet configured
at link could be morethan MTU. Hence there is a chance of receiving
plts morethan MTU which then gets DMA'ed into multiple buffers
and notified in a single CQE_RX. This patch treats such pkts as errors
and frees up receive buffers pointers back to hardware.

Also on the transmit side this patch sets SMQ MAXLEN to max value to avoid
HW length errors for the packets whose size > MTU, eg due to path MTU.

Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-vf: Link event notification support
Tomasz Duszynski [Fri, 20 Mar 2020 18:57:23 +0000 (00:27 +0530)]
octeontx2-vf: Link event notification support

VF shares physical link with PF. Admin function (AF) sends
notification to PF whenever a link change event happens. PF
has to forward the same notification to each of the enabled VF.

PF traps START/STOP_RX messages sent by VF to AF to keep track of
VF's enabled/disabled state.

Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-vf: Ethtool support
Tomasz Duszynski [Fri, 20 Mar 2020 18:57:22 +0000 (00:27 +0530)]
octeontx2-vf: Ethtool support

Added ethtool support for VF devices for
 - Driver stats, Tx/Rx perqueue stats
 - Set/show Rx/Tx queue count
 - Set/show Rx/Tx ring sizes
 - Set/show IRQ coalescing parameters
 - RSS configuration etc

It's the PF which owns the interface, hence VF
cannot display underlying CGX interface stats.
Except for this rest ethtool support reuses PF's
APIs.

Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-vf: Virtual function driver support
Tomasz Duszynski [Fri, 20 Mar 2020 18:57:21 +0000 (00:27 +0530)]
octeontx2-vf: Virtual function driver support

On OcteonTx2 silicon there two two types VFs, VFs that share the
physical link with their parent SR-IOV PF and the VFs which work
in pairs using internal HW loopback channels (LBK). Except for the
underlying Rx/Tx channel mapping from netdev functionality perspective
they are almost identical. This patch adds netdev driver support
for these VFs.

Unlike it's parent PF a VF cannot directly communicate with admin
function (AF) and it has to go through PF for the same. The mailbox
communication with AF works like 'VF <=> PF <=> AF'.

Also functionality wise VF and PF are identical, hence to avoid code
duplication PF driver's APIs are resued here for HW initialization,
packet handling etc etc ie almost everything. For VF driver to compile
as module exported few of the existing PF driver APIs.

Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-pf: Handle VF function level reset
Geetha sowjanya [Fri, 20 Mar 2020 18:57:20 +0000 (00:27 +0530)]
octeontx2-pf: Handle VF function level reset

When FLR is initiated for a VF (PCI function level reset),
the parent PF gets a interrupt. PF then sends a message to
admin function (AF), which then cleanups all resources attached
to that VF.

Also handled IRQs triggered when master enable bit is cleared
or set for VFs. This handler just clears the transaction pending
ie TRPEND bit.

Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoocteontx2-pf: Enable SRIOV and added VF mbox handling
Sunil Goutham [Fri, 20 Mar 2020 18:57:19 +0000 (00:27 +0530)]
octeontx2-pf: Enable SRIOV and added VF mbox handling

Added 'sriov_configure' to enable/disable virtual functions (VFs).
Also added handling of mailbox messages from these VFs.

Admin function (AF) is the only one with all priviliges to configure
HW, alloc resources etc etc, PFs and it's VFs have to request AF
via mbox for all their needs. But unlike PFs, their VFs cannot
send a mbox request directly. A VF shares a mailbox region with
it's parent PF, so VF sends a mailbox msg to PF and then PF forwards
it to AF. Then AF after processing sends response to PF which it
again forwards to VF.

This patch adds support for this 'VF <=> PF <=> AF' mailbox
communication.

Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Christina Jacob <cjacob@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'phy_check_downshift'
David S. Miller [Tue, 24 Mar 2020 04:09:47 +0000 (21:09 -0700)]
Merge branch 'phy_check_downshift'

Heiner Kallweit says:

====================
net: phy: add and use phy_check_downshift

So far PHY drivers have to check whether a downshift occurred to be
able to notify the user. To make life of drivers authors a little bit
easier move the downshift notification to phylib. phy_check_downshift()
compares the highest mutually advertised speed with the actual value
of phydev->speed (typically read by the PHY driver from a
vendor-specific register) to detect a downshift.

v2: Add downshift hint to phy_print_status().
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: aquantia: remove downshift warning now that phylib takes care
Heiner Kallweit [Fri, 20 Mar 2020 16:52:53 +0000 (17:52 +0100)]
net: phy: aquantia: remove downshift warning now that phylib takes care

Now that phylib notifies the user of a downshift we can remove
this functionality from the driver.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: marvell: remove downshift warning now that phylib takes care
Heiner Kallweit [Fri, 20 Mar 2020 16:52:10 +0000 (17:52 +0100)]
net: phy: marvell: remove downshift warning now that phylib takes care

Now that phylib notifies the user of a downshift we can remove
this functionality from the driver.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>