]> asedeno.scripts.mit.edu Git - linux.git/log
linux.git
5 years agoiwlwifi: mvm: d3: use struct_size() in kzalloc()
Gustavo A. R. Silva [Fri, 24 Aug 2018 01:15:40 +0000 (20:15 -0500)]
iwlwifi: mvm: d3: use struct_size() in kzalloc()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
int stuff;
        void *entry[];
};

instance = kzalloc(sizeof(struct foo) + sizeof(void *) * count, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL);

This issue was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: fix spelling mistake "Recieved" -> "Received"
Colin Ian King [Sun, 26 Aug 2018 22:31:17 +0000 (23:31 +0100)]
iwlwifi: mvm: fix spelling mistake "Recieved" -> "Received"

Trivial fix to spelling mistake in debug message.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: set TWT responder capability bit in 11AX SoftAP mode
Emmanuel Grumbach [Mon, 3 Sep 2018 11:32:48 +0000 (14:32 +0300)]
iwlwifi: mvm: set TWT responder capability bit in 11AX SoftAP mode

This is needed to test TWT requester on the client side.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: remove unused and wrong PHY_CFG_* macros
Luca Coelho [Thu, 30 Aug 2018 12:07:03 +0000 (15:07 +0300)]
iwlwifi: remove unused and wrong PHY_CFG_* macros

These macros are never used and are actually wrong, so it's very
confusing.  Remove them.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: wrt: add rt status and num of rx/tx fifos to dump
Shahar S Matityahu [Tue, 28 Aug 2018 13:15:55 +0000 (16:15 +0300)]
iwlwifi: wrt: add rt status and num of rx/tx fifos to dump

Add the rt status of the last assert or 0 if the dump collection was
not initiated by an assert.  Add the number of rx and tx fifos in use.
These fields are added to dump info lst file.

Signed-off-by: Shahar S Matityahu <shahar.s.matityahu@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: clean up SSN incrementation
Sara Sharon [Wed, 22 Aug 2018 07:50:49 +0000 (10:50 +0300)]
iwlwifi: mvm: clean up SSN incrementation

Sometimes, due to SCD bug, we need to start the queue with an
higher SSN. The queue allocation function currently increments
the SSN in the packet itself, but it is pointless, since this
value is overridden later by iwl_mvm_tx_mpdu with the value
from mvmsta->tid_data[tid].seq_number. Updating tid data is
sufficient.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: pcie: lock txq a bit later in reclaim code
Sara Sharon [Mon, 27 Aug 2018 11:07:20 +0000 (14:07 +0300)]
iwlwifi: pcie: lock txq a bit later in reclaim code

In reclaim code, we don't need to take the queue lock for
waking the queue. The code section is executed only when
the tx path is stopped, and since the reclaim path is not
executed in parallel to itself, no one can update the queue
pointers, and accessing them is safe without a lock.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: take station lock later in the code
Sara Sharon [Mon, 27 Aug 2018 10:56:41 +0000 (13:56 +0300)]
iwlwifi: mvm: take station lock later in the code

There is no need to lock mvm station for transport reclaim.
Move the locking down, after the reclaim.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: handle RX no data notification
Shaul Triebitz [Mon, 23 Jul 2018 08:40:30 +0000 (11:40 +0300)]
iwlwifi: mvm: handle RX no data notification

Handle RX no data notification, which is used for advertising NDP to
radiotap.

Signed-off-by: Golan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: split HE capabilities between AP and STA
Shaul Triebitz [Mon, 20 Aug 2018 13:07:26 +0000 (16:07 +0300)]
iwlwifi: split HE capabilities between AP and STA

HE capabilities differ between AP and STA.

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: make MVM and DVM depend on MAC80211
Luca Coelho [Mon, 27 Aug 2018 10:02:19 +0000 (13:02 +0300)]
iwlwifi: make MVM and DVM depend on MAC80211

It's not the iwlwifi module that depends on mac80211, but iwlmvm and
iwldvm.  To reflect this better, make MVM and DVM Kconfig options
depend on MAC80211 instead.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: fix setting HE ppe FW config
Naftali Goldstein [Wed, 22 Aug 2018 09:18:19 +0000 (12:18 +0300)]
iwlwifi: mvm: fix setting HE ppe FW config

The FW expects to get the ppe value for each NSS-BW pair in the same
format as in the he phy capabilities IE, which means that a value of 0
implies ppe should be used for BPSK (mcs 0). If there are no PPE
thresholds in the IE, or if for some NSS-RU pair there's no threshold
set for it (this could happen because it's a variable-sized field), it
means no PPE should not be used for that pair, so the value sent to FW
should be 7 which corresponds to "none".

Fixes: 514c30696fbc ("iwlwifi: add support for IEEE802.11ax")
Signed-off-by: Naftali Goldstein <naftali.goldstein@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: cleanup iwl_mvm_tx_skb_non_sta
Sara Sharon [Thu, 23 Aug 2018 10:40:16 +0000 (13:40 +0300)]
iwlwifi: mvm: cleanup iwl_mvm_tx_skb_non_sta

Make the coupling of station id and queue id clear. Group code
together. Remove outdated comment. Never use an undefined hw
queue as given from mac80211.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: remove pointless NULL assignment
Sara Sharon [Thu, 23 Aug 2018 08:03:17 +0000 (11:03 +0300)]
iwlwifi: mvm: remove pointless NULL assignment

Assigning mvmsta to be NULL when we are about to exit the
function is pointless. Remove it. Move the variable declaration
to the scope it is used.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: mvm: remove dead code
Sara Sharon [Thu, 23 Aug 2018 09:10:01 +0000 (12:10 +0300)]
iwlwifi: mvm: remove dead code

In iwl_mvm_tx_skb_non_sta(), in case of managed interface,
the AP station was supposed to be used for multicast frames
instead of the auxiliary station to avoid frames possibly
sent to an absent P2P GO. However, when moving to DQA mode,
this was broken as no valid queue was assigned. This is fixed
by a recent patch that directs all non-offchannel traffic to
ap station earlier in the TX path. However, the broken, and
now dead code, remained. Remove it.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoiwlwifi: pcie: use iwl_tx_cmd_gen2 and not iwl_tx_cmd
Sara Sharon [Mon, 20 Aug 2018 07:34:04 +0000 (10:34 +0300)]
iwlwifi: pcie: use iwl_tx_cmd_gen2 and not iwl_tx_cmd

When building AMSDU for gen2, code uses iwl_tx_cmd. The only
updated field is len, which is in the same location, so it
is not a bug. However, it is a bit confusing and error prone,
so change it.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoipw2x00: cleanup dead code
YueHaibing [Wed, 19 Dec 2018 03:00:13 +0000 (11:00 +0800)]
ipw2x00: cleanup dead code

Fix smatch warning:

drivers/net/wireless/intel/ipw2x00/ipw2100.c:5606
 shim__set_security() warn: always true condition '(sec->active_key <= 3) => (0-3 <= 3)'
drivers/net/wireless/intel/ipw2x00/ipw2200.c:10725
 shim__set_security() warn: always true condition '(sec->active_key <= 3) => (0-3 <= 3)'

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agocw1200: Fix concurrency use-after-free bugs in cw1200_hw_scan()
Jia-Ju Bai [Fri, 14 Dec 2018 03:55:21 +0000 (11:55 +0800)]
cw1200: Fix concurrency use-after-free bugs in cw1200_hw_scan()

The function cw1200_bss_info_changed() and cw1200_hw_scan() can be
concurrently executed.
The two functions both access a possible shared variable "frame.skb".

This shared variable is freed by dev_kfree_skb() in cw1200_upload_beacon(),
which is called by cw1200_bss_info_changed(). The free operation is
protected by a mutex lock "priv->conf_mutex" in cw1200_bss_info_changed().

In cw1200_hw_scan(), this shared variable is accessed without the
protection of the mutex lock "priv->conf_mutex".
Thus, concurrency use-after-free bugs may occur.

To fix these bugs, the original calls to mutex_lock(&priv->conf_mutex) and
mutex_unlock(&priv->conf_mutex) are moved to the places, which can
protect the accesses to the shared variable.

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: Fix access point mode
Stefan Wahren [Wed, 12 Dec 2018 19:20:06 +0000 (20:20 +0100)]
brcmfmac: Fix access point mode

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 <arend.vanspriel@broadcom.com>
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: fix roamoff=1 modparam
Stijn Tintel [Tue, 4 Dec 2018 18:29:05 +0000 (20:29 +0200)]
brcmfmac: fix roamoff=1 modparam

When the update_connect_param callback is set, nl80211 expects the flag
WIPHY_FLAG_SUPPORTS_FW_ROAM to be set as well. However, this flag is
only set when modparam roamoff=0, while the callback is set
unconditionally. Since commit 7f9a3e150ec7 this causes a warning in
wiphy_register, which breaks brcmfmac.

Disable the update_connect_param callback when roamoff=0 to fix this.

Fixes: 7f9a3e150ec7 ("nl80211: Update ERP info using NL80211_CMD_UPDATE_CONNECT_PARAMS")
Cc: Stable <stable@vger.kernel.org> # 4.19+
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agocw1200: convert to DEFINE_SHOW_ATTRIBUTE
Yangtao Li [Mon, 3 Dec 2018 13:31:45 +0000 (08:31 -0500)]
cw1200: convert to DEFINE_SHOW_ATTRIBUTE

Use DEFINE_SHOW_ATTRIBUTE macro to simplify the code.

Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoMerge tag 'iwlwifi-next-for-kalle-2018-12-14' of git://git.kernel.org/pub/scm/linux...
Kalle Valo [Fri, 14 Dec 2018 14:58:00 +0000 (16:58 +0200)]
Merge tag 'iwlwifi-next-for-kalle-2018-12-14' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi-next

Third batch of iwlwifi patches intended for v4.21

* Some cleanups in the HW configuration structures;
* More work on the debugging infrastructure;
* Fixes related to 802.11ax;
* Other cleanups and small fixes;

5 years agoiwlwifi: mvm: don't check if a pointer is set if it can't be unset
Emmanuel Grumbach [Tue, 14 Aug 2018 13:49:30 +0000 (16:49 +0300)]
iwlwifi: mvm: don't check if a pointer is set if it can't be unset

We used to have many versions of statistics notification
coming from the firmware. In one of the cleanup patches,
we forgot to clean the code that checks if data->general
is set. Since it is always set, remove the check.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: pcie: set cmd_len in the correct place
Sara Sharon [Mon, 20 Aug 2018 07:07:32 +0000 (10:07 +0300)]
iwlwifi: pcie: set cmd_len in the correct place

command len is set too early in the code, since when building
AMSDU, the size changes. This causes the byte count table to
have the wrong size.

Fixes: a0ec0169b7a9 ("iwlwifi: support new tx api")
Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: mvm: radiotap: remove UL_DL bit in HE TRIG_BASED
Shaul Triebitz [Wed, 25 Jul 2018 19:11:52 +0000 (22:11 +0300)]
iwlwifi: mvm: radiotap: remove UL_DL bit in HE TRIG_BASED

UL_DL is irrelevant to HE TRIG_BASED PPDU.

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: mvm: add L-SIG length to radiotap
Shaul Triebitz [Mon, 2 Jul 2018 15:16:48 +0000 (17:16 +0200)]
iwlwifi: mvm: add L-SIG length to radiotap

We may have the L-SIG length depending on the phy_data info type;
add it to radiotap when we do.

Move getting the phy_data out one layer up and the info type into
it so we can use this data more generically. We need to call the
iwl_mvm_rx_he() function for other reasons as well, so can't just
combine all of that into something like iwl_mvm_parse_phy_data().

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: mvm: change PHY data RX for HE radiotap
Shaul Triebitz [Mon, 2 Jul 2018 12:35:26 +0000 (14:35 +0200)]
iwlwifi: mvm: change PHY data RX for HE radiotap

The firmware changed the PHY data API, so follow suit.
Some data is now available even for HT/VHT frames, so
the info type in the metadata was changed. This change
isn't backwards compatible, but
 1) the firmware with the old API was never released;
 2) the only overlap in the info type field is from the
    old type of TB to the new of HT, so this basically
    just means that with older FW and newer driver the
    data will be considered missing.

While at it, remove the extra code to set the LTF syms
corresponding to the streams and use the data from the
device instead - we don't really need this in any case
other than when we have it from the device.

As the new API gives use the spatial reuse 1-4 fields
for trigger-based PPDUs, also expose that to radiotap.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: pcie: fix erroneous print
Sara Sharon [Thu, 16 Aug 2018 10:01:09 +0000 (13:01 +0300)]
iwlwifi: pcie: fix erroneous print

When removing the driver, the following flow can happen:
1. host command is in progress, for example at index 68.
2. RX interrupt is received with the response.
3. Before it is processed, the remove flow kicks in, and
   calls iwl_pcie_txq_unmap. The function cleans all DMA,
   and promotes the read pointer to 69.
4. RX thread proceeds with the processing, and is calling
   iwl_pcie_cmdq_reclaim, which will print this error:
   iwl_pcie_cmdq_reclaim: Read index for DMA queue txq id (0),
   index 4 is out of range [0-256] 69 69.

Detect this situation, and avoid the print. Change it to
warning while at it, to make such issues more noticeable
in the future.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: mvm: force TCM re-evaluation on TCM resume
Avraham Stern [Wed, 15 Aug 2018 09:23:39 +0000 (12:23 +0300)]
iwlwifi: mvm: force TCM re-evaluation on TCM resume

When traffic load is not low or low latency is active, TCM schedules
re-evaluation work so in case traffic stops TCM will detect that
traffic load has become low or that low latency is no longer active.
However, if TCM is paused when the re-evaluation work runs, it does
not re-evaluate and the re-evaluation work is no longer scheduled.
As a result, TCM will not indicate that low latency is no longer
active or that traffic load is low when traffic stops.

Fix this by forcing TCM re-evaluation when TCM is resumed in case
low latency is active or traffic load is not low.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: remove all unnecessary occurrences of nvm_calib_ver
Luca Coelho [Thu, 16 Aug 2018 07:01:00 +0000 (10:01 +0300)]
iwlwifi: remove all unnecessary occurrences of nvm_calib_ver

Now that nvm_calib_ver is not checked in opmodes other than dvm, we
can remove it from all irrelevant configurations.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: move iwl_nvm_check_version() into dvm
Luca Coelho [Thu, 16 Aug 2018 06:23:08 +0000 (09:23 +0300)]
iwlwifi: move iwl_nvm_check_version() into dvm

This function is only half-used by mvm (i.e. only the nvm_version part
matters, since the calibration version is irrelevant), so it's
pointless to export it from iwlwifi.  If mvm uses this function, it
has the additional complexity of setting the calib version to a bogus
value on all cfg structs.

To avoid this, move the function to dvm and make a simple comparison
of the nvm_version in mvm instead.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: remove useless NVM_HW_SECTION_NUM_FAMILY_* macros
Luca Coelho [Thu, 16 Aug 2018 06:01:31 +0000 (09:01 +0300)]
iwlwifi: remove useless NVM_HW_SECTION_NUM_FAMILY_* macros

These macros are useless because each one of them is used only once
and the element they are assigned to is already pretty clear about
what they mean, "nvm_hw_section_num".

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: don't define OTP_LOW_IMAGE_SIZE per family, but per size
Luca Coelho [Thu, 16 Aug 2018 05:53:10 +0000 (08:53 +0300)]
iwlwifi: don't define OTP_LOW_IMAGE_SIZE per family, but per size

Using OTP_LOW_IMAGE_SIZE_FAMILY_8000/9000/22000 only obfuscates the
actual values, since these 3 are the same.  Redefine the values per
size so it's easier to understand and compare the different
configurations.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: mvm: activate apply points
Sara Sharon [Wed, 13 Jun 2018 12:22:48 +0000 (15:22 +0300)]
iwlwifi: mvm: activate apply points

Call the previously introduced apply points entry
point when reaching an apply point.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: add support for triggering ini triggers
Sara Sharon [Thu, 21 Jun 2018 06:42:12 +0000 (09:42 +0300)]
iwlwifi: add support for triggering ini triggers

Add support for ini triggers.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: dbg: dump data according to the new ini TLVs
Sara Sharon [Wed, 13 Jun 2018 12:24:13 +0000 (15:24 +0300)]
iwlwifi: dbg: dump data according to the new ini TLVs

When ini TLVs are loaded, dump data according to the
stored configuration.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agoiwlwifi: move common 9000 cfg parameters to common macro
Luca Coelho [Thu, 9 Aug 2018 10:51:08 +0000 (13:51 +0300)]
iwlwifi: move common 9000 cfg parameters to common macro

The ht_params, nvm_ver, nvm_calib_ver and max_ht_ampdu_exponent
elements in 9000 devices are always the same.  Move them to the common
macro.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
5 years agortlwifi: rtl8821ae: remove set but not used variable 'type' in translate_rx_signal_stuff
YueHaibing [Thu, 6 Dec 2018 12:06:08 +0000 (12:06 +0000)]
rtlwifi: rtl8821ae: remove set but not used variable 'type' in translate_rx_signal_stuff

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

drivers/net/wireless/realtek/rtlwifi/rtl8821ae/trx.c: In function 'translate_rx_signal_stuff':
drivers/net/wireless/realtek/rtlwifi/rtl8821ae/trx.c:309:6: warning:
 variable 'type' set but not used [-Wunused-but-set-variable]

It never used since commit 21e4b0726dc6 ("rtlwifi: rtl8821ae: Move driver from
staging to regular tree")

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agortl818x: fix potential use after free
Pan Bian [Thu, 29 Nov 2018 10:48:10 +0000 (18:48 +0800)]
rtl818x: fix potential use after free

entry is released via usb_put_urb just after calling usb_submit_urb.
However, entry is used if the submission fails, resulting in a use after
free bug. The patch fixes this.

Signed-off-by: Pan Bian <bianpan2016@163.com>
ACKed-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agob43: remove set but not used variables 'tx_pwr_state,tmp2'
YueHaibing [Wed, 5 Dec 2018 07:26:06 +0000 (07:26 +0000)]
b43: remove set but not used variables 'tx_pwr_state,tmp2'

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

drivers/net/wireless/broadcom/b43/phy_n.c: In function 'b43_nphy_op_recalc_txpower':
drivers/net/wireless/broadcom/b43/phy_n.c:5898:7: warning:
 variable 'tx_pwr_state' set but not used [-Wunused-but-set-variable]

drivers/net/wireless/broadcom/b43/phy_n.c:6047:20: warning:
 variable 'tmp2' set but not used [-Wunused-but-set-variable]

'tx_pwr_state' never used since introduction in commit
8615eb2870f1 ("b43: N-PHY: support setting custom TX power")

'tmp2' not used any more since commit c002831a07dd ("b43: N-PHY: use helper
for checking IPA")

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: fix false-positive -Wmaybe-unintialized warning
Arnd Bergmann [Mon, 10 Dec 2018 20:55:37 +0000 (21:55 +0100)]
brcmfmac: fix false-positive -Wmaybe-unintialized warning

When CONFIG_NO_AUTO_INLINE is set, we get a false-postive warning
for the brcmf_fw_request_nvram_done() function, after gcc figures
out that brcmf_fw_nvram_from_efi() might not set the 'data_len'
variable, but fails to notice that it always returns NULL:

drivers/net/wireless/broadcom/brcm80211/brcmfmac/firmware.c: In function 'brcmf_fw_request_nvram_done':
drivers/net/wireless/broadcom/brcm80211/brcmfmac/firmware.c:560:11: error: 'data_len' may be used uninitialized in this function [-Werror=maybe-uninitialized]

Mark it 'inline' to force gcc to understand this.

Fixes: ce2e6db554fa ("brcmfmac: Add support for getting nvram contents from EFI variables")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: disable command decode in sdio_aos
Wright Feng [Wed, 21 Nov 2018 07:53:52 +0000 (07:53 +0000)]
brcmfmac: disable command decode in sdio_aos

AOS is a part of the SDIOD core that becomes active when the rest of
SDIOD is sleeping to keep SDIO bus alive responding to reduced set of
commands.

Transaction between AOS and SDIOD is not protected, and if cmd 52 is
received in AOS and in the middle of response state changed from AOS to
SDIOD, response is corrupted and it causes to SDIO Host controller to
hang.

Command decode for below chips are disabled in this commit:
 - 4339
 - 4345
 - 4354
 - 4373

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Wright Feng <wright.feng@cypress.com>
Signed-off-by: Double Lo <double.lo@cypress.com>
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>
5 years agobrcmfmac: 4373 save-restore support
Chi-Hsien Lin [Wed, 21 Nov 2018 07:53:50 +0000 (07:53 +0000)]
brcmfmac: 4373 save-restore support

Use chipcommon sr_control0 register to check 4373 sr support.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: update 43012 F2 watermark setting to fix DMA Error during UDP RX Traffic
Naveen Gupta [Wed, 21 Nov 2018 07:53:49 +0000 (07:53 +0000)]
brcmfmac: update 43012 F2 watermark setting to fix DMA Error during UDP RX Traffic

The number of words that the read FIFO has to contain except
the end of frame before sends data back to the host.
Max watermark = (512B - 2* (BurstLength))/4 =
(512 - 128)/4 = 384/4 = 0x60
so if burst length (i.e. BurstLength = 64) is increased,
watermark has to be reduced. This is the optimal setting for this chip.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Naveen Gupta <naveen.gupta@cypress.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: allow GCI core enumuration
Chi-Hsien Lin [Wed, 21 Nov 2018 07:53:48 +0000 (07:53 +0000)]
brcmfmac: allow GCI core enumuration

GCI core is needed for ULP operation. Allow GCI core enumuration with
below changes:
 - Allow GCI to be added to core list even when it doesn't have a wrapper.
 - Allow 8K address space size.
 - Don't overwrite the address value when an additional size descriptor
   is in place.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: add support for CYW43012 SDIO chipset
Chi-Hsien Lin [Wed, 21 Nov 2018 07:53:47 +0000 (07:53 +0000)]
brcmfmac: add support for CYW43012 SDIO chipset

CYW43012 is a 1x1 802.11a/b/g/n Dual-Band HT20, 256-QAM/Turbo QAM. It
is an Ultra Low Power WLAN+BT combo chip.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Praveen Babu C <praveen.chandran@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: set SDIO F1 MesBusyCtrl for CYW4373
Madhan Mohan R [Wed, 21 Nov 2018 07:53:45 +0000 (07:53 +0000)]
brcmfmac: set SDIO F1 MesBusyCtrl for CYW4373

Along with F2 watermark (existing) configuration, F1 MesBusyCtrl
should be enabled & sdio device RX FIFO watermark should be
configured to avoid overflow errors.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
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>
5 years agobrcmfmac: set F2 watermark to 256 for 4373
Wright Feng [Wed, 21 Nov 2018 07:53:44 +0000 (07:53 +0000)]
brcmfmac: set F2 watermark to 256 for 4373

We got SDIO_CRC_ERROR with 4373 on SDR104 when doing bi-directional
throughput test. Enable watermark to 256 to guarantee the operation
stability.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
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>
5 years agobrcmfmac: add 4354 raw pcie device id
Winnie Chang [Wed, 21 Nov 2018 07:53:42 +0000 (07:53 +0000)]
brcmfmac: add 4354 raw pcie device id

Add the raw 4354 PCIe device ID for unprogrammed Cypress boards.

Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Signed-off-by: Winnie Chang <winnie.chang@cypress.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: handle compressed tx status signal
Chung-Hsien Hsu [Mon, 5 Nov 2018 05:52:05 +0000 (05:52 +0000)]
brcmfmac: handle compressed tx status signal

Firmware inform the driver about tx status by normal tx status signal
or compressed tx status signal. This patch adds support to handle the
compressed tx status signal.

Signed-off-by: Chung-Hsien Hsu <stanley.hsu@cypress.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Signed-off-by: Wright Feng <wright.feng@cypress.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: enable frameburst mode in default firmware setting
Wright Feng [Mon, 5 Nov 2018 05:51:59 +0000 (05:51 +0000)]
brcmfmac: enable frameburst mode in default firmware setting

The frameburst feature can enable per-packet framebursting in firmware
side and get higher TX throughput in High Throughput(HT) mode. To enhance
TX throughput, we enable frameburst mode in default firmware setting.

Signed-off-by: Wright Feng <wright.feng@cypress.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agobrcmfmac: add credit numbers updating support
Wright Feng [Mon, 5 Nov 2018 05:51:54 +0000 (05:51 +0000)]
brcmfmac: add credit numbers updating support

The credit numbers are static and tunable per chip in firmware side.
However the credit number may be changed that is based on packet pool
length and will send BRCMF_E_FIFO_CREDIT_MAP event to notify host driver
updates the credit numbers during interface up.
The purpose of this patch is making host driver has ability of updating
the credit numbers when receiving the BRCMF_E_FIFO_CREDIT_MAP event.

Signed-off-by: Wright Feng <wright.feng@cypress.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agomwifiex: add NL80211_STA_INFO_RX_BITRATE support
Brian Norris [Fri, 7 Dec 2018 00:12:49 +0000 (16:12 -0800)]
mwifiex: add NL80211_STA_INFO_RX_BITRATE support

Comparing the existing TX_BITRATE parsing code (in
mwifiex_parse_htinfo()) with the RX bitrate histograms in debugfs.c, it
appears that the rxpd_rate and rxpd_htinfo fields have the same format.
At least, they give reasonable results when I parse them this way.

So this patch adds support for RX_BITRATE to our station info dump.

Along the way, I add legacy bitrate parsing into the same function,
using the debugfs code (mwifiex_histogram_read() and
mwifiex_adjust_data_rate()) as reference.

Additionally, to satisfy the requirements of
NL80211_STA_INFO_RX_BITRATE, I skip logging the bitrate of multicast
packets. This shouldn't add a lot of overhead to the RX path, as there
are already several similar 802.3 header checks in this same codepath.
We can also bias the branch behavior to favor unicast, as that's the
common performance-sensitive case.

I'd consider this support somewhat experimental, as I have zero
documentation from Marvell. But the existing driver code gives me good
reason to think this is correct.

I've tested this on a few different 802.11{a,b,g,n,ac} networks, and the
reported bitrates look good to me.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agomwifiex: refactor mwifiex_parse_htinfo() for reuse
Brian Norris [Fri, 7 Dec 2018 00:12:48 +0000 (16:12 -0800)]
mwifiex: refactor mwifiex_parse_htinfo() for reuse

This function converts some firmware-specific parameters into cfg80211
'rate_info' structures. It currently assumes it's dealing only with TX
bitrate, but the RX bitrate looks to be the same, so refactor this
function to be reusable.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agomwifiex: debugfs: correct histogram spacing, formatting
Brian Norris [Mon, 3 Dec 2018 22:26:55 +0000 (14:26 -0800)]
mwifiex: debugfs: correct histogram spacing, formatting

Currently, snippets of this file look like:

rx rates (in Mbps): 0=1M   1=2M2=5.5M  3=11M   4=6M   5=9M  6=12M
7=18M  8=24M  9=36M  10=48M  11=54M12-27=MCS0-15(BW20) 28-43=MCS0-15(BW40)
44-53=MCS0-9(VHT:BW20)54-63=MCS0-9(VHT:BW40)64-73=MCS0-9(VHT:BW80)
...
noise_flr[--96dBm] = 22
noise_flr[--95dBm] = 149
noise_flr[--94dBm] = 9
noise_flr[--93dBm] = 2

We're missing some spaces, and we're adding a minus sign ('-') on values
that are already negative signed integers.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agomwifiex: fix potential NULL dereference and use after free
Pan Bian [Thu, 29 Nov 2018 10:25:19 +0000 (18:25 +0800)]
mwifiex: fix potential NULL dereference and use after free

There are two defects: (1) passing a NULL bss to
mwifiex_save_hidden_ssid_channels will result in NULL dereference,
(2) using bss after dropping the reference to it via cfg80211_put_bss.
To fix them, the patch moves the buggy code to the branch that bss is
not NULL and puts it before cfg80211_put_bss.

Signed-off-by: Pan Bian <bianpan2016@163.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
5 years agoMerge tag 'mt76-for-kvalo-2018-11-30' of https://github.com/nbd168/wireless
Kalle Valo [Thu, 13 Dec 2018 14:02:19 +0000 (16:02 +0200)]
Merge tag 'mt76-for-kvalo-2018-11-30' of https://github.com/nbd168/wireless

first batch of mt76 patches for 4.21

* use the same firmware for mt76x2e and mt76x2u
* mt76x2 fixes
* mt76x0 fixes
* mt76x0e survey support
* more unification between mt76x2 and mt76x0
* mt76x0e AP mode support
* mt76x0e DFS support
* rework and fix tx status handling for mt76x0 and mt76x2

kvalo: fixed a conflict in drivers/net/wireless/mediatek/mt76/mt76x2/pci_main.c
and a build problem in drivers/net/wireless/mediatek/mt76/mt76x02_util.c

5 years agonet: marvell: convert to DEFINE_SHOW_ATTRIBUTE
Yangtao Li [Mon, 3 Dec 2018 14:34:11 +0000 (09:34 -0500)]
net: marvell: convert to DEFINE_SHOW_ATTRIBUTE

Use DEFINE_SHOW_ATTRIBUTE macro to simplify the code.

Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: qca_spi: convert to DEFINE_SHOW_ATTRIBUTE
Yangtao Li [Mon, 3 Dec 2018 14:26:38 +0000 (09:26 -0500)]
net: qca_spi: convert to DEFINE_SHOW_ATTRIBUTE

Use DEFINE_SHOW_ATTRIBUTE macro to simplify the code.

Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: convert to DEFINE_SHOW_ATTRIBUTE
Yangtao Li [Mon, 3 Dec 2018 14:22:09 +0000 (09:22 -0500)]
net: stmmac: convert to DEFINE_SHOW_ATTRIBUTE

Use DEFINE_SHOW_ATTRIBUTE macro to simplify the code.

Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: convert to DEFINE_SHOW_ATTRIBUTE
Yangtao Li [Mon, 3 Dec 2018 14:13:35 +0000 (09:13 -0500)]
nfp: convert to DEFINE_SHOW_ATTRIBUTE

Use DEFINE_SHOW_ATTRIBUTE macro to simplify the code.

Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mlx4_core-cleanups'
David S. Miller [Tue, 4 Dec 2018 00:34:32 +0000 (16:34 -0800)]
Merge branch 'mlx4_core-cleanups'

Tariq Toukan says:

====================
mlx4_core cleanups

This patchset by Erez contains cleanups to the mlx4_core driver.

Patch 1 replaces -EINVAL with -EOPNOTSUPP for unsupported operations.
Patch 2 fixes some coding style issues.

Series generated against net-next commit:
97e6c858a26e net: usb: aqc111: Initialize wol_cfg with memset in aqc111_suspend
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx4_core: Fix several coding style errors
Erez Alfasi [Sun, 2 Dec 2018 15:40:26 +0000 (17:40 +0200)]
net/mlx4_core: Fix several coding style errors

Fix 3 checkpatch errors within mlx4/main.c:
- Unnecessary mlx4_debug_level global variable initialization to 0.
- Prohibited space before comma.
- Whitespaces instead of tab.

Signed-off-by: Erez Alfasi <ereza@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx4_core: Fix return codes of unsupported operations
Erez Alfasi [Sun, 2 Dec 2018 15:40:25 +0000 (17:40 +0200)]
net/mlx4_core: Fix return codes of unsupported operations

Functions __set_port_type and mlx4_check_port_params returned
-EINVAL while the proper return code is -EOPNOTSUPP as a
result of an unsupported operation. All drivers should generate
this and all users should check for it when detecting an
unsupported functionality.

Signed-off-by: Erez Alfasi <ereza@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: Also request modules for C45 IDs
Jose Abreu [Sun, 2 Dec 2018 15:33:14 +0000 (16:33 +0100)]
net: phy: Also request modules for C45 IDs

Logic of phy_device_create() requests PHY modules according to PHY ID
but for C45 PHYs we use different field for the IDs.

Let's also request the modules for these IDs.

Changes from v1:
- Only request C22 modules if C45 are not present (Andrew)

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Joao Pinto <joao.pinto@synopsys.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'octeontx2-next'
David S. Miller [Tue, 4 Dec 2018 00:23:09 +0000 (16:23 -0800)]
Merge branch 'octeontx2-next'

Jerin Jacob says:

====================
octeontx2-af: NIX and NPC enhancements

This patchset is a continuation to earlier submitted four patch
series to add a new driver for Marvell's OcteonTX2 SOC's
Resource virtualization unit (RVU) admin function driver.

1. octeontx2-af: Add RVU Admin Function driver
   https://www.spinics.net/lists/netdev/msg528272.html
2. octeontx2-af: NPA and NIX blocks initialization
   https://www.spinics.net/lists/netdev/msg529163.html
3. octeontx2-af: NPC parser and NIX blocks initialization
   https://www.spinics.net/lists/netdev/msg530252.html
4. octeontx2-af: NPC MCAM support and FLR handling
   https://www.spinics.net/lists/netdev/msg534392.html

This patch series adds support for below

NPC block:
- Add NPC(mkex) profile support for various Key extraction configurations

NIX block:
- Enable dynamic RSS flow key algorithm configuration
- Enhancements on Rx checksum and error checks
- Add support for Tx packet marking support
- TL1 schedule queue allocation enhancements
- Add LSO format configuration mbox
- VLAN TPID configuration
- Skip multicast entry init for broadcast tables

v2:

- Rename FLOW_* to NIX_FLOW_* to avoid serious global namespace collisions,
as we have various FLOW_* definitions coming from
include/uapi/linux/pkt_cls.h for example.(David Miller)
- Pack the arguments of rvu_get_tl1_schqs() function
as 80 columns allows.(David Miller)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Enable mkex profile
Vamsi Attunuru [Sun, 2 Dec 2018 12:47:49 +0000 (18:17 +0530)]
octeontx2-af: Enable mkex profile

The following set of NPC registers allow the driver to configure NPC
to generate different key value schemes to compare against packet
payload in MCAM search.

NPC_AF_INTF(0..1)_KEX_CFG
NPC_AF_KEX_LDATA(0..1)_FLAGS_CFG
NPC_AF_INTF(0..1)_LID(0..7)_LT(0..15)_LD(0..1)_CFG
NPC_AF_INTF(0..1)_LDATA(0..1)_FLAGS(0..15)_CFG

Currently, the AF driver populates these registers to
configure the default values to address the most common
use cases such as key generation for channel number + DMAC.

The secure firmware stores different configuration
value of these registers to enable different NPC use case
along with the name for the lookup.

Patch loads profile binary from secure firmware over
the exiting CGX mailbox interface and apply the profile.

AF driver shall fall back to the default configuration
in case of any errors.

The AF consumer driver can know the selected profile
on response to NPC_GET_KEX_CFG mailbox by introducing
mkex_pfl_name in the struct npc_get_kex_cfg_rsp.

Signed-off-by: Vamsi Attunuru <vamsi.attunuru@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Add LSO format configuration mailbox
Nithin Dabilpuram [Sun, 2 Dec 2018 12:47:48 +0000 (18:17 +0530)]
octeontx2-af: Add LSO format configuration mailbox

NIX_AF_LSO_FORMAT(0..31)_FIELD(0..7) register enables an SW defined
means to define LSO packet modification formats.

0..31 works as an index to choose the algorithm, On success, the mailbox
returns the index to the client of chosen LSO algorithm selection.
This index will be used in configuring the transmit descriptors.

Add mailbox interface to dynamically reserve and configure LSO format.

This commit also fixes 'sizem1' for NIX_LSOALG_TCP_FLAGS
to '1' i.e 2 Bytes.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Add L3 and L4 packet verification mailbox
Vidhya Raman [Sun, 2 Dec 2018 12:47:47 +0000 (18:17 +0530)]
octeontx2-af: Add L3 and L4 packet verification mailbox

Adds mailbox support for L4 checksum verification
and L3 and L4 length verification configuration.

Signed-off-by: Vidhya Raman <vraman@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Configure VLAN TPIDs
Nithin Dabilpuram [Sun, 2 Dec 2018 12:47:46 +0000 (18:17 +0530)]
octeontx2-af: Configure VLAN TPIDs

Setup TPID's for vlan0 and vlan1 for Tx VLAN insertion offloads.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Add support for Tx packet marking
Krzysztof Kanas [Sun, 2 Dec 2018 12:47:45 +0000 (18:17 +0530)]
octeontx2-af: Add support for Tx packet marking

NIX_AF_MARK_FORMAT(0..127)_CTL register enables an SW defined
means to mark/insert various data in the packet based on
final packet color from traffic shaping HW.

0..127 works as an index to choose the algorithm. On success,
the mailbox returns the index to the client.

Add NIX_MARK_FORMAT_CFG mailbox which reserves mark format based on
tuple (offset, y_mask, y_val, r_mask, r_val)

If the tuple is requested again for mark format that was already
reserved, then it will be reused. If not it will reserve a new entry
if space is available.

Also on AF init commonly used marker format such as VLAN DEI, IPv4
ECN, IPv4 DSCP are reserved for AF consumers.

Signed-off-by: Krzysztof Kanas <kkanas@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Enable RSS with promiscuous mode
Vamsi Attunuru [Sun, 2 Dec 2018 12:47:44 +0000 (18:17 +0530)]
octeontx2-af: Enable RSS with promiscuous mode

This patch adds support for enabling RSS in promiscuous mode
if RSS is already requested by the AF client.

Signed-off-by: Vamsi Attunuru <vamsi.attunuru@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Define all NIX_AF_RX_DEF_* registers
Jerin Jacob [Sun, 2 Dec 2018 12:47:43 +0000 (18:17 +0530)]
octeontx2-af: Define all NIX_AF_RX_DEF_* registers

In order to support all NIX specific valid length errors and
checksum errors on Rx, Update all NIX_AF_RX_DEF_* registers.

Also sorted all registers in HRM definition order.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Enable inner IPv4 checksum and its error code
Jerin Jacob [Sun, 2 Dec 2018 12:47:42 +0000 (18:17 +0530)]
octeontx2-af: Enable inner IPv4 checksum and its error code

This patch enables the inner IPv4 checksum and
defines the error code for Rx inner and outer checksum errors.
Setting ERRCODE as 1 so that CQE descriptor can be embedded
valid checksum error code and the driver can interpret
checksum error as ERRLEV = LID + 1 and ERRCODE = 1.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Allow freeing single TLx Tx schedule queue
Nithin Dabilpuram [Sun, 2 Dec 2018 12:47:41 +0000 (18:17 +0530)]
octeontx2-af: Allow freeing single TLx Tx schedule queue

The default behavior was to free all the TLx Tx schedule
queues. This patch adds support for freeing a single Tx
schedule queue if TXSCHQ_FREE_ALL flag is not set.

Signed-off-by: Krzysztof Kanas <kkanas@marvell.com>
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Restrict TL1 allocation and configuration
Nithin Dabilpuram [Sun, 2 Dec 2018 12:47:40 +0000 (18:17 +0530)]
octeontx2-af: Restrict TL1 allocation and configuration

TL1 is the root node in the scheduling hierarchy and
it is a global resource with a limited number.

This patch introduces restriction and validation on
the allocation of the TL1 nodes for the effective resource
sharing across the AF consumers.

- Limit TL1 allocation to 2 per lmac.
  One could be for the normal link and one for IEEE802.3br
  express link (Express Send DMA).
  Effectively all the VF's of an RVU PF(lmac) share the two TL1 schqs.
- TL1 cannot be freed once allocated.
- Allow VF's to only apply default config to TL1 if not
  already applied. PF's can always overwrite the TL1 config.
- Consider NIX_AQ_INSTOP_WRITE while validating txschq
  when sq.ena is set.

Signed-off-by: Krzysztof Kanas <kkanas@marvell.com>
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Add support for runtime RSS algo index reservation
Jerin Jacob [Sun, 2 Dec 2018 12:47:39 +0000 (18:17 +0530)]
octeontx2-af: Add support for runtime RSS algo index reservation

Introduced reserve_flowkey_alg_idx()to reserve RSS algorithm index,
it would internally use set_flowkey_fields() to generate fields
based on the flow key dynamically.

On AF driver init, it would reserve a predefined set RSS algo indexes,
which will be available all the time for all the AF driver consumers.
The leftover algo indexes can be reserved at runtime through
exiting nix_rss_flowkey_cfg mailbox message.

The NIX_FLOW_KEY_TYPE_PORT is removed from predefined a set of RSS flow
type as it is not used by any consumer.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Add support for dynamic flow cfg to RSS field generation
Jerin Jacob [Sun, 2 Dec 2018 12:47:38 +0000 (18:17 +0530)]
octeontx2-af: Add support for dynamic flow cfg to RSS field generation

Introduce state-based algorithm to convert the flow_key value
to RSS algo field used by NIX_AF_RX_FLOW_KEY_ALGX_FIELDX register.

The outer `for loop` goes over _all_ protocol field and the following
variables depict the state machine forward progress logic.

a) keyoff_marker - Enabled when hash byte length needs to be accounted
in field->key_offset update.
b) field_marker - Enabled when a new field needs to be selected.
c) group_member - Enabled when a protocol is part of a group.

This would remove the existing hard coding and enable to add
new protocol support seamlessly.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Add response for RSS flow key cfg message
Jerin Jacob [Sun, 2 Dec 2018 12:47:37 +0000 (18:17 +0530)]
octeontx2-af: Add response for RSS flow key cfg message

Added response for nix_rss_flowkey_cfg message to return
selected RSS algorithm index.

The FLOW_KEY_TYPE* definition is part of the mbox message and
it will be used by the other consumers of AF driver hence moving to mbox.h.

Also renamed FLOW_* definitions to NIX_FLOW_* to avoid global
name space collisions, as we have various coming from
include/uapi/linux/pkt_cls.h for example.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoocteontx2-af: Skip NIXLF check for bcast MCE entry
Sunil Goutham [Sun, 2 Dec 2018 12:47:36 +0000 (18:17 +0530)]
octeontx2-af: Skip NIXLF check for bcast MCE entry

At the time of initial broadcast packet replication table init,
NIXLFs are not yet attached to PF_FUNCs. Hence skipped checking
NIXLF while submitting MCE entry init instruction to NIX admin queue.

Also did a minor cleanup while installing bcast match entry in
packet parser unit i.e NPC.

Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'udp-msg_zerocopy'
David S. Miller [Mon, 3 Dec 2018 23:58:33 +0000 (15:58 -0800)]
Merge branch 'udp-msg_zerocopy'

Willem de Bruijn says:

====================
udp msg_zerocopy

Enable MSG_ZEROCOPY for udp sockets

Patch 1/3 is the main patch, a rework of RFC patch
  http://patchwork.ozlabs.org/patch/899630/
  more details in the patch commit message

Patch 2/3 is an optimization to remove a branch from the UDP hot path
  and refcount_inc/refcount_dec_and_test pair when zerocopy is used.
  This used to be included in the first patch in v2.

Patch 3/3 runs the already existing udp zerocopy tests
  as part of kselftest

See also recent Linux Plumbers presentation
  https://linuxplumbersconf.org/event/2/contributions/106/attachments/104/128/willemdebruijn-lpc2018-udpgso-presentation-20181113.pdf

Changes:
  v1 -> v2
    - Fixup reverse christmas tree violation
  v2 -> v3
    - Split refcount avoidance optimization into separate patch
      - Fix refcount leak on error in fragmented case
        (thanks to Paolo Abeni for pointing this one out!)
      - Fix refcount inc on zero
  v3 -> v4
    - Move skb_zcopy_set below the only kfree_skb that might cause
      a premature uarg destroy before skb_zerocopy_put_abort
      - Move the entire skb_shinfo assignment block, to keep that
cacheline access in one place
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoselftests: extend zerocopy tests to udp
Willem de Bruijn [Fri, 30 Nov 2018 20:32:41 +0000 (15:32 -0500)]
selftests: extend zerocopy tests to udp

Both msg_zerocopy and udpgso_bench have udp zerocopy variants.
Exercise these as part of the standard kselftest run.

With udp, msg_zerocopy has no control channel. Ensure that the
receiver exits after the sender by accounting for the initial
delay in starting them (in msg_zerocopy.sh).

Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoudp: elide zerocopy operation in hot path
Willem de Bruijn [Fri, 30 Nov 2018 20:32:40 +0000 (15:32 -0500)]
udp: elide zerocopy operation in hot path

With MSG_ZEROCOPY, each skb holds a reference to a struct ubuf_info.
Release of its last reference triggers a completion notification.

The TCP stack in tcp_sendmsg_locked holds an extra ref independent of
the skbs, because it can build, send and free skbs within its loop,
possibly reaching refcount zero and freeing the ubuf_info too soon.

The UDP stack currently also takes this extra ref, but does not need
it as all skbs are sent after return from __ip(6)_append_data.

Avoid the extra refcount_inc and refcount_dec_and_test, and generally
the sock_zerocopy_put in the common path, by passing the initial
reference to the first skb.

This approach is taken instead of initializing the refcount to 0, as
that would generate error "refcount_t: increment on 0" on the
next skb_zcopy_set.

Changes
  v3 -> v4
    - Move skb_zcopy_set below the only kfree_skb that might cause
      a premature uarg destroy before skb_zerocopy_put_abort
      - Move the entire skb_shinfo assignment block, to keep that
        cacheline access in one place

Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoudp: msg_zerocopy
Willem de Bruijn [Fri, 30 Nov 2018 20:32:39 +0000 (15:32 -0500)]
udp: msg_zerocopy

Extend zerocopy to udp sockets. Allow setting sockopt SO_ZEROCOPY and
interpret flag MSG_ZEROCOPY.

This patch was previously part of the zerocopy RFC patchsets. Zerocopy
is not effective at small MTU. With segmentation offload building
larger datagrams, the benefit of page flipping outweights the cost of
generating a completion notification.

tools/testing/selftests/net/msg_zerocopy.sh after applying follow-on
test patch and making skb_orphan_frags_rx same as skb_orphan_frags:

    ipv4 udp -t 1
    tx=191312 (11938 MB) txc=0 zc=n
    rx=191312 (11938 MB)
    ipv4 udp -z -t 1
    tx=304507 (19002 MB) txc=304507 zc=y
    rx=304507 (19002 MB)
    ok
    ipv6 udp -t 1
    tx=174485 (10888 MB) txc=0 zc=n
    rx=174485 (10888 MB)
    ipv6 udp -z -t 1
    tx=294801 (18396 MB) txc=294801 zc=y
    rx=294801 (18396 MB)
    ok

Changes
  v1 -> v2
    - Fixup reverse christmas tree violation
  v2 -> v3
    - Split refcount avoidance optimization into separate patch
      - Fix refcount leak on error in fragmented case
        (thanks to Paolo Abeni for pointing this one out!)
      - Fix refcount inc on zero
      - Test sock_flag SOCK_ZEROCOPY directly in __ip_append_data.
        This is needed since commit 5cf4a8532c99 ("tcp: really ignore
MSG_ZEROCOPY if no SO_ZEROCOPY") did the same for tcp.

Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'wireless-drivers-next-for-davem-2018-11-30' of git://git.kernel.org/pub...
David S. Miller [Mon, 3 Dec 2018 23:44:27 +0000 (15:44 -0800)]
Merge tag 'wireless-drivers-next-for-davem-2018-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for 4.21

First set of patches for 4.21. Most notable here is support for
Quantenna's QSR1000/QSR2000 chipsets and more flexible ways to provide
nvram files for brcmfmac.

Major changes:

brcmfmac

* add support for first trying to get a board specific nvram file

* add support for getting nvram contents from EFI variables

qtnfmac

* use single PCIe driver for all platforms and rename
  Kconfig option CONFIG_QTNFMAC_PEARL_PCIE to CONFIG_QTNFMAC_PCIE

* add support for QSR1000/QSR2000 (Topaz) family of chipsets

ath10k

* add support for WCN3990 firmware crash recovery

* add firmware memory dump support for QCA4019

wil6210

* add firmware error recovery while in AP mode

ath9k

* remove experimental notice from dynack feature

iwlwifi

* PCI IDs for some new 9000-series cards

* improve antenna usage on connection problems

* new firmware debugging infrastructure

* some more work on 802.11ax

* improve support for multiple RF modules with 22000 devices

cordic

* move cordic macros and defines to a public header file

* convert brcmsmac and b43 to fully use cordic library
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'davinci_emac-read-the-MAC-address-from-nvmem'
David S. Miller [Mon, 3 Dec 2018 23:40:30 +0000 (15:40 -0800)]
Merge branch 'davinci_emac-read-the-MAC-address-from-nvmem'

Bartosz Golaszewski says:

====================
davinci_emac: read the MAC address from nvmem

This series is part of a bigger series that aims at removing the platform
data structure from the at24 EEPROM driver[1].

We provide a generalized version of of_get_nvmem_mac_address(), switch the
only user of the of_ variant to using it, remove the previous
implementation and use the new routine in the davinci_emac driver.

[1] https://lkml.org/lkml/2018/11/13/884
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: davinci_emac: use nvmem_get_mac_address()
Bartosz Golaszewski [Fri, 30 Nov 2018 08:21:00 +0000 (09:21 +0100)]
net: davinci_emac: use nvmem_get_mac_address()

All DaVinci boards still supported in board files now define nvmem
cells containing the MAC address. We want to stop using the setup
callback from at24 so the MAC address for those users will no longer
be provided over platform data. If we didn't get a valid MAC in pdata,
try nvmem before resorting to a random MAC.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoof: net: kill of_get_nvmem_mac_address()
Bartosz Golaszewski [Fri, 30 Nov 2018 08:20:59 +0000 (09:20 +0100)]
of: net: kill of_get_nvmem_mac_address()

We've switched all users to nvmem_get_mac_address(). Remove the now
dead code.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: cadence: switch to using nvmem_get_mac_address()
Bartosz Golaszewski [Fri, 30 Nov 2018 08:20:58 +0000 (09:20 +0100)]
net: cadence: switch to using nvmem_get_mac_address()

We now have a generalized helper routine to read the MAC address from
nvmem which takes struct device as argument. The nvmem subsystem will
then try device tree first before all other potential providers.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: provide nvmem_get_mac_address()
Bartosz Golaszewski [Fri, 30 Nov 2018 08:20:57 +0000 (09:20 +0100)]
net: ethernet: provide nvmem_get_mac_address()

We already have of_get_nvmem_mac_address() but some non-DT systems want
to read the MAC address from NVMEM too. Implement a generalized routine
that takes struct device as argument.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agorhashtable: detect when object movement between tables might have invalidated a lookup
NeilBrown [Thu, 29 Nov 2018 23:26:50 +0000 (10:26 +1100)]
rhashtable: detect when object movement between tables might have invalidated a lookup

Some users of rhashtables might need to move an object from one table
to another -  this appears to be the reason for the incomplete usage
of NULLS markers.

To support these, we store a unique NULLS_MARKER at the end of
each chain, and when a search fails to find a match, we check
if the NULLS marker found was the expected one.  If not, the search
may not have examined all objects in the target bucket, so it is
repeated.

The unique NULLS_MARKER is derived from the address of the
head of the chain.  As this cannot be derived at load-time the
static rhnull in rht_bucket_nested() needs to be initialised
at run time.

Any caller of a lookup function must still be prepared for the
possibility that the object returned is in a different table - it
might have been there for some time.

Note that this does NOT provide support for other uses of
NULLS_MARKERs such as allocating with SLAB_TYPESAFE_BY_RCU or changing
the key of an object and re-inserting it in the same table.
These could only be done safely if new objects were inserted
at the *start* of a hash chain, and that is not currently the case.

Signed-off-by: NeilBrown <neilb@suse.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'hns3-ethtool-dump'
David S. Miller [Mon, 3 Dec 2018 23:26:34 +0000 (15:26 -0800)]
Merge branch 'hns3-ethtool-dump'

Salil Mehta says:

====================
Adds VF/PF PCIe reg dump(ethtool -d) support to HNS3 driver

This patchset adds VF/PF PCIe register dump support to HNS3 VF and PF
driver using "ethtool -d" command.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: Adds support to dump(using ethool-d) PCIe regs in HNS3 PF driver
Jian Shen [Thu, 29 Nov 2018 16:40:59 +0000 (16:40 +0000)]
net: hns3: Adds support to dump(using ethool-d) PCIe regs in HNS3 PF driver

This patch adds support to dump PF PCIe registers using ethtool -d
for HNS3 PF Driver.

Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: Support "ethtool -d" for HNS3 VF driver
Jian Shen [Thu, 29 Nov 2018 16:40:58 +0000 (16:40 +0000)]
net: hns3: Support "ethtool -d" for HNS3 VF driver

This patch adds "ethtool -d" support for HNS3 VF Driver.

Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: improve generic EEE ethtool functions
Heiner Kallweit [Tue, 27 Nov 2018 21:30:14 +0000 (22:30 +0100)]
net: phy: improve generic EEE ethtool functions

So far the two functions consider neither member eee_enabled nor
eee_active. Therefore network drivers have to do this in some kind
of glue code. I think this can be avoided.

Getting EEE parameters:
When not advertising any EEE mode, we can't consider EEE to be enabled.
Therefore interpret "EEE enabled" as "we advertise at least one EEE
mode". It's similar with "EEE active": interpret it as "EEE modes
advertised by both link partner have at least one mode in common".

Setting EEE parameters:
If eee_enabled isn't set, don't advertise any EEE mode and restart
aneg if needed to switch off EEE. If eee_enabled is set and
data->advertised is empty (e.g. because EEE was disabled), advertise
everything we support as default. This way EEE can easily switched
on/off by doing ethtool --set-eee <if> eee on/off, w/o any additional
parameters.

The changes to both functions shouldn't break any existing user.
Once the changes have been applied, at least some users can be
simplified.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'VXLAN-underlay-VRF'
David S. Miller [Mon, 3 Dec 2018 22:16:14 +0000 (14:16 -0800)]
Merge branch 'VXLAN-underlay-VRF'

Alexis Bauvin says:

====================
net: Add VRF support for VXLAN underlay

v6 -> v7:
- proper locking for device in udp_tunnel following Sabrina Dubroca's advice

v5 -> v6:
- remove automatic rebinding patch following Roopa Prabhu's advice

v4 -> v5:
- move test script to its own patch (6/6)
- add schematic for test script
- apply David Ahern comments to the test script

v3 -> v4:
- rename vxlan_is_in_l3mdev_chain to netdev_is_upper master
- move it to net/core/dev.c
- make it return bool instead of int
- check if remote_ifindex is zero before resolving the l3mdev
- add testing script

v2 -> v3:
- fix build when CONFIG_NET_IPV6 is off
- fix build "unused l3mdev_master_upper_ifindex_by_index" build error with some
  configs

v1 -> v2:
- move vxlan_get_l3mdev from vxlan driver to l3mdev driver as
  l3mdev_master_upper_ifindex_by_index
- vxlan: rename variables named l3mdev_ifindex to ifindex

v0 -> v1:
- fix typos

We are trying to isolate the VXLAN traffic from different VMs with VRF as shown
in the schemas below:

+-------------------------+   +----------------------------+
| +----------+            |   |     +------------+         |
| |          |            |   |     |            |         |
| | tap-red  |            |   |     |  tap-blue  |         |
| |          |            |   |     |            |         |
| +----+-----+            |   |     +-----+------+         |
|      |                  |   |           |                |
|      |                  |   |           |                |
| +----+---+              |   |      +----+----+           |
| |        |              |   |      |         |           |
| | br-red |              |   |      | br-blue |           |
| |        |              |   |      |         |           |
| +----+---+              |   |      +----+----+           |
|      |                  |   |           |                |
|      |                  |   |           |                |
|      |                  |   |           |                |
| +----+--------+         |   |     +--------------+       |
| |             |         |   |     |              |       |
| |  vxlan-red  |         |   |     |  vxlan-blue  |       |
| |             |         |   |     |              |       |
| +------+------+         |   |     +-------+------+       |
|        |                |   |             |              |
|        |           VRF  |   |             |          VRF |
|        |           red  |   |             |         blue |
+-------------------------+   +----------------------------+
         |                                  |
         |                                  |
 +---------------------------------------------------------+
 |       |                                  |              |
 |       |                                  |              |
 |       |         +--------------+         |              |
 |       |         |              |         |              |
 |       +---------+  eth0.2030   +---------+              |
 |                 |  10.0.0.1/24 |                        |
 |                 +-----+--------+                    VRF |
 |                       |                            green|
 +---------------------------------------------------------+
                         |
                         |
                    +----+---+
                    |        |
                    |  eth0  |
                    |        |
                    +--------+

iproute2 commands to reproduce the setup:

ip link add green type vrf table 1
ip link set green up
ip link add eth0.2030 link eth0 type vlan id 2030
ip link set eth0.2030 master green
ip addr add 10.0.0.1/24 dev eth0.2030
ip link set eth0.2030 up

ip link add blue type vrf table 2
ip link set blue up
ip link add br-blue type bridge
ip link set br-blue master blue
ip link set br-blue up
ip link add vxlan-blue type vxlan id 2 local 10.0.0.1 dev eth0.2030 \
 port 4789
ip link set vxlan-blue master br-blue
ip link set vxlan-blue up
ip link set tap-blue master br-blue
ip link set tap-blue up

ip link add red type vrf table 3
ip link set red up
ip link add br-red type bridge
ip link set br-red master red
ip link set br-red up
ip link add vxlan-red type vxlan id 3 local 10.0.0.1 dev eth0.2030 \
 port 4789
ip link set vxlan-red master br-red
ip link set vxlan-red up
ip link set tap-red master br-red
ip link set tap-red up

We faced some issue in the datapath, here are the details:

* Egress traffic:
The vxlan packets are sent directly to the default VRF because it's where the
socket is bound, therefore the traffic has a default route via eth0. the
workaround is to force this traffic to VRF green with ip rules.

* Ingress traffic:
When receiving the traffic on eth0.2030 the vxlan socket is unreachable from
VRF green. The workaround is to enable *udp_l3mdev_accept* sysctl, but
this breaks isolation between overlay and underlay: packets sent from
blue or red by e.g. a guest VM will be accepted by the socket, allowing
injection of VXLAN packets from the overlay.

This patch series fixes the issues describe above by allowing VXLAN socket to be
bound to a specific VRF device therefore looking up in the correct table.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotest/net: Add script for VXLAN underlay in a VRF
Alexis Bauvin [Mon, 3 Dec 2018 09:54:41 +0000 (10:54 +0100)]
test/net: Add script for VXLAN underlay in a VRF

This script tests the support of a VXLAN underlay in a non-default VRF.

It does so by simulating two hypervisors and two VMs, an extended L2
between the VMs with the hypervisors as VTEPs with the underlay in a
VRF, and finally by pinging the two VMs.

It also tests that moving the underlay from a VRF to another works when
down/up the VXLAN interface.

Signed-off-by: Alexis Bauvin <abauvin@scaleway.com>
Reviewed-by: Amine Kherbouche <akherbouche@scaleway.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: Amine Kherbouche <akherbouche@scaleway.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agovxlan: add support for underlay in non-default VRF
Alexis Bauvin [Mon, 3 Dec 2018 09:54:40 +0000 (10:54 +0100)]
vxlan: add support for underlay in non-default VRF

Creating a VXLAN device with is underlay in the non-default VRF makes
egress route lookup fail or incorrect since it will resolve in the
default VRF, and ingress fail because the socket listens in the default
VRF.

This patch binds the underlying UDP tunnel socket to the l3mdev of the
lower device of the VXLAN device. This will listen in the proper VRF and
output traffic from said l3mdev, matching l3mdev routing rules and
looking up the correct routing table.

When the VXLAN device does not have a lower device, or the lower device
is in the default VRF, the socket will not be bound to any interface,
keeping the previous behaviour.

The underlay l3mdev is deduced from the VXLAN lower device
(IFLA_VXLAN_LINK).

+----------+                         +---------+
|          |                         |         |
| vrf-blue |                         | vrf-red |
|          |                         |         |
+----+-----+                         +----+----+
     |                                    |
     |                                    |
+----+-----+                         +----+----+
|          |                         |         |
| br-blue  |                         | br-red  |
|          |                         |         |
+----+-----+                         +---+-+---+
     |                                   | |
     |                             +-----+ +-----+
     |                             |             |
+----+-----+                +------+----+   +----+----+
|          |  lower device  |           |   |         |
|   eth0   | <- - - - - - - | vxlan-red |   | tap-red | (... more taps)
|          |                |           |   |         |
+----------+                +-----------+   +---------+

Signed-off-by: Alexis Bauvin <abauvin@scaleway.com>
Reviewed-by: Amine Kherbouche <akherbouche@scaleway.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: Amine Kherbouche <akherbouche@scaleway.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agol3mdev: add function to retreive upper master
Alexis Bauvin [Mon, 3 Dec 2018 09:54:39 +0000 (10:54 +0100)]
l3mdev: add function to retreive upper master

Existing functions to retreive the l3mdev of a device did not walk the
master chain to find the upper master. This patch adds a function to
find the l3mdev, even indirect through e.g. a bridge:

+----------+
|          |
| vrf-blue |
|          |
+----+-----+
     |
     |
+----+-----+
|          |
| br-blue  |
|          |
+----+-----+
     |
     |
+----+-----+
|          |
|   eth0   |
|          |
+----------+

This will properly resolve the l3mdev of eth0 to vrf-blue.

Signed-off-by: Alexis Bauvin <abauvin@scaleway.com>
Reviewed-by: Amine Kherbouche <akherbouche@scaleway.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: Amine Kherbouche <akherbouche@scaleway.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoudp_tunnel: add config option to bind to a device
Alexis Bauvin [Mon, 3 Dec 2018 09:54:38 +0000 (10:54 +0100)]
udp_tunnel: add config option to bind to a device

UDP tunnel sockets are always opened unbound to a specific device. This
patch allow the socket to be bound on a custom device, which
incidentally makes UDP tunnels VRF-aware if binding to an l3mdev.

Signed-off-by: Alexis Bauvin <abauvin@scaleway.com>
Reviewed-by: Amine Kherbouche <akherbouche@scaleway.com>
Tested-by: Amine Kherbouche <akherbouche@scaleway.com>
Signed-off-by: David S. Miller <davem@davemloft.net>