mwifiex: uninit wakeup info in the error handling

We inited wakeup info at the beginning of mwifiex_add_card, so we need
to uninit it in the error handling.

It's much the same as what we did in:
36908c4 mwifiex: uninit wakeup info when removing device

Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
Reviewed-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: fix compile warning of unused variable

We got a compile warning shows below:

drivers/net/wireless/marvell/mwifiex/sdio.c: In function
'mwifiex_sdio_remove':
drivers/net/wireless/marvell/mwifiex/sdio.c:377:6: warning: variable
'ret' set but not used [-Wunused-but-set-variable]

Per the code, it didn't check if mwifiex_sdio_read_fw_status
finish successfully. We should at least check the return of
mwifiex_sdio_read_fw_status, otherwise the following check of
firmware_stat and adapter->mfg_mode is pointless as the device
is probably dead.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Reviewed-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: drop num CPU notice

This print isn't very useful. It's also different between
mwifiex_add_card() and mwifiex_reinit_sw(), and I'd like to consolidate
them eventually.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: keep mwifiex_cancel_pending_ioctl() static

It has some scary comments about "only being called" from the timeout
handler, so let's help keep it that way.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: pcie: remove unnecessary 'pdev' check

'card->dev' is initialized once and is never cleared. Drop the
unnecessary "safety" check, as it simply obscures things, and we don't
do this check everywhere (and therefore it's not really "safe").

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: pcie: disable device DMA before unmapping/freeing buffers

In testing the mwifiex reset code path, I've noticed KASAN complaining
about some "overwritten poison values" in our RX buffer descriptors.
Because KASAN didn't notice this at the time of a CPU write, this seems
to suggest that the device is writing to this memory.

This makes a little sense, because when resetting, we don't necessarily
expect the device to be responsive, so we don't have a chance to disable
everything cleanly.

We can at least take the precaution of disabling DMA for the device
though, and in my testing that seems to clear up this particular issue.

This patch reorders the removal path so that we disable the device
*before* releasing our last PCIe buffers, and it clears/sets the bus
master feature from the PCI device when resetting.

Along the way, remove the insufficient (and confusing) error path in
mwifiex_pcie_up_dev() (it doesn't unwind things well enough, and it
doesn't propagate its errors upward anyway).

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: debugfs: allow card_reset() to cancel things

The card_reset() implementation should be setting our state flags and
cancelling commands for us (i.e., in mwifiex_shutdown_drv()), so let's
not do it here.

Also, this debugfs file is useful for testing and debugging the reset
feature, so we shouldn't do extra preparatory steps here, as that might
cause different reset behavior, which could either cause new bugs or
paper over existing ones that this debug feature should otherwise help
us catch.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: pcie: unify MSI-X / non-MSI-X interrupt process

After removing the interrupt loop in commit 5d5ddb5e0d9b ("mwifiex:
pcie: don't loop/retry interrupt status checks"), there is practically
zero difference between mwifiex_process_pcie_int() (which handled legacy
PCI interrupts and MSI interrupts) and mwifiex_process_msix_int() (which
handled MSI-X interrupts). Let's add the one relevant line to
mwifiex_process_pcie_int() and kill the copy-and-paste.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: pcie: remove unnecessary masks

After removing the interrupt loop in commit 5d5ddb5e0d9b ("mwifiex:
pcie: don't loop/retry interrupt status checks"), we don't need to keep
track of the cleared interrupts (actually, we didn't need to do that
before, but we *really* don't need to now).

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: drop 'add_tail' param from mwifiex_insert_cmd_to_pending_q()

It's always called with 'true' -- we only determine it 'false' locally
within this function. So drop the parameter.

Also, this should be 'bool' (since we use true/false), not 'u32'.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: don't open-code ARRAY_SIZE()

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: utilize netif_tx_{wake,stop}_all_queues()

We're open-coding these. Just use the helpers.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: make mwifiex_free_cmd_buffer() return void

It doesn't fail.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: fix misnomers in mwifiex_free_lock_list()

Despite the name (and meticulous comments), this function frees no
memory and does not touch any locks. All it does is "delete" the list
heads -- which just means they'll be dangling, and we'll need to re-init
them if we use them again.

It seems like this code would work OK as a sort of canary for using the
list after we've torn everything down, so it's fine to keep the code;
let's just get the name and comments to match what's actually happening.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: ensure "disable auto DS" struct is initialized

The .idle_time field *should* be unused, but technically, we're allowing
unitialized stack garbage to pass all the way through to the firmware
host command. Let's zero it out instead.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: fixup init_channel_scan_gap error case

In reading through _mwifiex_fw_dpc(), I noticed that after we've
registered our wiphy, we still have error paths that don't free it back
up. Let's do that.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: don't short-circuit netdev notifiers on interface deletion

When we leave the delete interface function, there are still netdev
hooks that might try to process the device. We're short-circuiting some
of that by changing the interface type and clearing ieee80211_ptr. This
means we skip NETDEV_UNREGISTER_FINAL in cfg80211. Fortunately, that is
currently a no-op.

We don't need most of the cleanup here anyway:

 * the connection state will get (un)set as part of the disconnect
   process (which cfg80211 already initiates for us)
 * the interface type doesn't actually need to be cleared at all (it'll
   trigger a WARN_ON() in cfg80211 if we do)
 * the iee80211_ptr isn't really "ours" to clear anyway

So stop resetting those 3 things.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: unregister wiphy before freeing resources

It's possible for some control interfaces (e.g., scans, set freq) to be
active after we've stopped our main work queue and the netif TX queues.
These don't get completely shut out until we've unregistered the wdevs
and wiphy.

So let's only free command buffers and poison our lists after
wiphy_unregister().

This resolves various use-after-free issues seen when resetting the
device.

Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: re-register wiphy across reset

In general, it's helpful to use the same code for device removal as for
device reset, as this tends to have fewer bugs. Let's move the wiphy
unregistration code into the common reset and removal code.

In particular, it's very hard to properly handle the reset sequence when
something fails. Currently, if mwifiex_reinit_sw() fails, we've failed
to unregister the associated wiphy, and so running something as simple
as "iw phy" can trigger an OOPS, as the wiphy still has hooks back into
freed mwifiex data structures. For example, KASAN complained:

[... see reset fail for other reasons ...]
[ 1184.821158] mwifiex_pcie 0000:01:00.0: info: dnld wifi firmware from 174948 bytes
[ 1186.870914] mwifiex_pcie 0000:01:00.0: info: FW download over, size 608396 bytes
[ 1187.685990] mwifiex_pcie 0000:01:00.0: WLAN FW is active
[ 1187.692673] mwifiex_pcie 0000:01:00.0: cmd_wait_q terminated: -512
[ 1187.699075] mwifiex_pcie 0000:01:00.0: info: _mwifiex_fw_dpc: unregister device
[ 1187.713476] mwifiex: Failed to bring up adapter: -5
[ 1187.718644] mwifiex_pcie 0000:01:00.0: reinit failed: -5

[... run `iw phy` ...]
[ 1212.902419] ==================================================================
[ 1212.909806] BUG: KASAN: use-after-free in mwifiex_cfg80211_get_antenna+0x54/0xfc [mwifiex] at addr ffffffc0ad1a8028
[ 1212.920246] Read of size 1 by task iw/3127
[...]
[ 1212.934946] page dumped because: kasan: bad access detected
[...]
[ 1212.950665] Call trace:
[ 1212.953148] [<ffffffc00020a69c>] dump_backtrace+0x0/0x190
[ 1212.958572] [<ffffffc00020a96c>] show_stack+0x20/0x28
[ 1212.963648] [<ffffffc0005ce18c>] dump_stack+0xa4/0xcc
[ 1212.968723] [<ffffffc0003c4430>] kasan_report+0x378/0x500
[ 1212.974140] [<ffffffc0003c3358>] __asan_load1+0x44/0x4c
[ 1212.979462] [<ffffffbffc2e8360>] mwifiex_cfg80211_get_antenna+0x54/0xfc [mwifiex]
[ 1212.987131] [<ffffffbffc084fc4>] nl80211_send_wiphy+0x75c/0x2de0 [cfg80211]
[ 1212.994246] [<ffffffbffc094f60>] nl80211_dump_wiphy+0x32c/0x438 [cfg80211]
[ 1213.001149] [<ffffffc000ab6404>] genl_lock_dumpit+0x48/0x64
[ 1213.006746] [<ffffffc000ab3474>] netlink_dump+0x178/0x398
[ 1213.012171] [<ffffffc000ab3d18>] __netlink_dump_start+0x1bc/0x260
[...]

This all goes away if we just tear down the wiphy on the way down, and
set it back up if/when we bring the device back up.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: pcie: don't allow cmd buffer reuse after reset

In rogue cases (due to other bugs) it's possible we try to process an
old command response *after* resetting the device. This could trigger a
double-free (or the SKB can get reallocated elsewhere...causing other
memory corruptions) in mwifiex_pcie_process_cmd_complete().

For safety (and symmetry) let's always NULL out the command buffer as we
free it up. We're already doing this for the command response buffer.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: reset interrupt status across device reset

When resetting the device, we might have queued up interrupts that
didn't get a chance to finish processing. We really don't need to handle
them at this point; we just want to make sure they don't cause us to try
to process old commands from before the device was reset.

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: reunite copy-and-pasted remove/reset code

When PCIe FLR code was added, it explicitly copy-and-pasted much of
mwifiex_remove_card() into mwifiex_shutdown_sw(). This is unnecessary,
as almost all of the code should be reused.

Let's reunite what we can for now.

The only functional changes for now:

 * call netif_device_detach() in the remove() code path -- this wasn't
   done before, but it really should be a no-op, when the device is
   getting totally unregistered soon anyway

 * call the ->down_dev() driver callback only after we've finished all
   SW teardown -- this should have no significant effect, since the only
   user (pcie.c) does very minimal work there, and it doesn't matter
   that we reorder this

Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: fix static checker warning

u32 pointer is changed to u16 and filled the value. Problem is
solved by using local temporary variable. Below static checker
warning was reported.

drivers/net/wireless/rsi/rsi_91x_usb.c:400 rsi_usb_master_reg_read()
        warn: passing casted pointer 'value' to 'rsi_usb_reg_read()'
32 vs 16.

Fixes: b97e9b94ad75c ("rsi: Add new host interface operations")
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: check length before USB read/write register

These checks are required. Otherwise we may end up getting
memory corruption if invalid length is passed.

Fixes: b97e9b94ad75c ("rsi: Add new host interface operations")
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: use macro for allocating USB buffer

4 bytes is fixed size for reading or writing USB register. We
will use a macro instead of hardcoding this.

Fixes: b97e9b94ad75c ("rsi: Add new host interface operations")
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: regulatory enhancements

Below regulatory changes are included this patch
* Country code is saved as it will be used in bgscan.
* Region codes are mapped according to RSI region codes.
* Radar flag settings are moved under the check if 5GHZ band
  is enabled.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Send rx filter frame to device when interface is down

When interface is down it is better to stop all RX packets to
host. 0xffff will block all packets to host.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Remove internal header from Tx status skb

Device specific descriptor for each TX packet is prepared on top of
skb data address. This descriptor should be pulled out before
indicating the TX status to mac80211.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: update tx command frame block/unblock data

TX command frame block/unblock data is modified to
use common descriptor structure.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: block/unblock data queues as per connection status

Data queues should be unblocked after station add notify frame
is sent and should be blocked after station delete notify is sent.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: update autorate request command frame

When Short Gaurd Interval is enabled bit 9 is set in rate. Otherwise
it should not be set. Added missing 'else' case in this patch.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: set_key enhancements

Broadcast bit to be set for broadcast and multicast packets.
For remove_key, tx and rx mic need not be filled.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: update set_key command frame

TX command frame set_key is modified to use common descriptor
structure.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: update vap capabilities command frame

VAP capablities frame configured to device is modified to use
common descriptor structure.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: update set_channel command frame

TX command frame set_channel is modified to use common
descriptor structure. Also DFS channel indication to firmware
is added in the descriptor for dfs channels configuration.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Update baseband RF programming frame

Baseband RF programming frame configured to device is modified
to use common descriptor structure.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Update aggregation parameters command frame

AMPDU aggregation parameters frame configured to device is
modified to use common descriptor structure.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Update peer notify command frame

TX command frame peer notify is updated to use common descriptor
structure. MPDU density value added to the frame.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: remove unnecessary check for 802.11 management packet

The function rsi_mgmt_pkt_to_core() is for passing the 802.11
management frames to mac80211. So, it is unnecessary to check
again for the frame type 802.11 management in this function.
It can be checked before passing to this function.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Update in tx command frame radio capabilities

Radio capabilities frame is updated to use common descriptor
structure. Also, few changes to this frame is done like hardware
queues are increase to 12 from 8, default channel number is included.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: immediate wakeup bit and priority for TX command packets

For all TX command packets immediate wakeup bit needs to be
set in descriptor. This will make sure device will wakeup if it
is in any sleep state. Priority of the packet is also set.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: add common structures needed for command packets

All internal management packets (command packets) use some common
fields and some packet specific fields for packet descriptors.
This patch adds some common structures which are needed for all
command packets.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Rename mutex tx_rxlock to the tx_lock.

We have now added separate lock for Rx. This lock is used to protect
tx path only

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: use separate mutex lock for receive thread

Deadlock issue is observed during our stress tests. The root
cause for the issue is same lock is used between tx and rx threads.

This patch adds a separate mutex lock for rx thread to resolve
the problem.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: SDIO Rx packet processing enhancement

Newer firmware sends information about number of blocks through
interrupt only. We don't need to read extra register for this.
This patch adds needed driver changes for this enhancment. The
change here is backward compatible

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: Optimise sdio claim and release host

SDIO host is already claimed in our interrupt handler. Some lower
level APIs claims host while performing SDIO read or write operations.
Let's use sdio_irq_task variable to check if we are in interrupt
context and claim/release the host accordingly.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: rename variable in_sdio_litefi_irq

The variable in_sdio_litefi_irq is renamed to sdio_irq_task,
as it is more relevant here.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: separate function for data packet descriptor

Data packet descriptor preparation code is moved a separate
function.

Signed-off-by: Pavani Muthyala <pavani.muthyala@redpinesignals.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: data packet descriptor enhancements

This patch covers some enhancements in data packet descriptor
preparation especially for EAPOL, multicast/broadcast packets.

Signed-off-by: Pavani Muthyala <pavani.muthyala@redpinesignals.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: data packet descriptor code cleanup

Currently this descriptor is prepared with the help of __le16
pointer. This patch makes use of a structure to prepare the
descriptor in a cleaner way.

Signed-off-by: Pavani Muthyala <pavani.muthyala@redpinesignals.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: separate function for management packet descriptor

Management descriptor preparation is move to a separate function
as it will be called from a different context in upcoming patches.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: management frame descriptor preparation cleanup

Currently this descriptor is prepared with the help of __le16
pointer. This patch makes use of a structure to prepare the
descriptor in a cleaner way.

Signed-off-by: Pavani Muthyala <pavani.muthyala@redpinesignals.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: set immediate wakeup bit

immediate wakeup bit is set while sending internal management
frame to the firmware.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: choose correct endpoint based on queue.

Till now only management packets were handled. Let's enhance
the logic for choosing endpoint to accommodate other packets.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: rename USB endpoint macros

These endpoints are for WLAN and BT protocols.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: correct the logic of deriving queue number

Maximum valid queue number is 0x5. So anding with 0x7 should be
ok here.

Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: USB tx headroom cleanup

USB headroom is added while submitting the data to URB as per
firmware's requirement. This logic is moved to rsi_usb_card_write() so
that caller need not worry about it.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: card reset for USB interface

We need to reset the chip in teardown path so that it can work
next time when driver is loaded. This patch adds support for this
reset configuration for USB.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: correct SDIO disconnect path handling

Sometimes it's observed that we get interrupt/Rx frame when device is
already detached from mac80211. In this case couple of error messages
are displayed in dmesg log. This patch corrects the order so that
disconnection will happen cleanly

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: chip reset for SDIO interface

We need to reset the chip in teardown path so that it can work
next time when driver is loaded. This patch adds support for
this reset configuration for SDIO.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: fix sdio card reset problem

card reset is not working with recent kernels. Using
host->card->ocr instead of host->ocr_avail solved the problem.

Signed-off-by: Karun Eagalapati <karun256@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: changes in eeprom read frame

EEPROM read frame is sent during device initialization to read mac address.
The format of the frame is modified in firmware to include eeprom length
and offset. This frame does not return firmware version now. Also same
frame is sent again to read rf type and band information.

Signed-off-by: Prameela Rani Garnepudi <prameela.j04cs@gmail.com>
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rsi: use BUILD_BUG_ON check for fsm_state

Whenever new fsm_state enum element is added, fsm_state array
also needs to be updated. If this change is missed, we may end
up doing invalid access in array. BUILD_BUG_ON check will help
to avoid this problem.

Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

Merge ath-next from git://git./linux/kernel/git/kvalo/ath.git

ath.git patches for 4.14. Major changes:

ath10k

* preparation for wcn3990 support

wl1251: add a missing spin_lock_init()

This fixes the following kernel warning:

  [ 5668.771453] BUG: spinlock bad magic on CPU#0, kworker/u2:3/9745
  [ 5668.771850]  lock: 0xce63ef20, .magic: 00000000, .owner: <none>/-1,
  .owner_cpu: 0
  [ 5668.772277] CPU: 0 PID: 9745 Comm: kworker/u2:3 Tainted: G        W
  4.12.0-03002-gec979a4-dirty #40
  [ 5668.772796] Hardware name: Nokia RX-51 board
  [ 5668.773071] Workqueue: phy1 wl1251_irq_work
  [ 5668.773345] [<c010c9e4>] (unwind_backtrace) from [<c010a274>]
  (show_stack+0x10/0x14)
  [ 5668.773803] [<c010a274>] (show_stack) from [<c01545a4>]
  (do_raw_spin_lock+0x6c/0xa0)
  [ 5668.774230] [<c01545a4>] (do_raw_spin_lock) from [<c06ca578>]
  (_raw_spin_lock_irqsave+0x10/0x18)
  [ 5668.774658] [<c06ca578>] (_raw_spin_lock_irqsave) from [<c048c010>]
  (wl1251_op_tx+0x38/0x5c)
  [ 5668.775115] [<c048c010>] (wl1251_op_tx) from [<c06a12e8>]
  (ieee80211_tx_frags+0x188/0x1c0)
  [ 5668.775543] [<c06a12e8>] (ieee80211_tx_frags) from [<c06a138c>]
  (__ieee80211_tx+0x6c/0x130)
  [ 5668.775970] [<c06a138c>] (__ieee80211_tx) from [<c06a3dbc>]
  (ieee80211_tx+0xdc/0x104)
  [ 5668.776367] [<c06a3dbc>] (ieee80211_tx) from [<c06a4af0>]
  (__ieee80211_subif_start_xmit+0x454/0x8c8)
  [ 5668.776824] [<c06a4af0>] (__ieee80211_subif_start_xmit) from
  [<c06a4f94>] (ieee80211_subif_start_xmit+0x30/0x2fc)
  [ 5668.777343] [<c06a4f94>] (ieee80211_subif_start_xmit) from
  [<c0578848>] (dev_hard_start_xmit+0x80/0x118)
  ...

by adding the missing spin_lock_init().

Reported-by: Pavel Machek <pavel@ucw.cz>
Cc: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

bcma: gpio: Correct number of GPIOs for BCM53573

Broadcom BCM53573 SoCs actually have 32 GPIOs, and not 16.

Fixes: 3f37ec79dd21 ("bcma: support BCM53573 series of wireless SoCs")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

rtlwifi: kfree entry until after entry->bssid has been accessed

The current code kfree's entry and then dereferences it by accessing
entry->bssid.  Avoid the dereference-after-free by moving the kfree
after the access to entry->bssid.

Detected by CoverityScan, CID#1448600 ("Read from pointer after free")

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

mwifiex: correct channel stat buffer overflows

mwifiex records information about various channels as it receives scan
information. It does this by appending to a buffer that was sized
to the max number of supported channels on any band, but there are
numerous problems:

(a) scans can return info from more than one band (e.g., both 2.4 and 5
    GHz), so the determined "max" is not large enough
(b) some firmware appears to return multiple results for a given
    channel, so the max *really* isn't large enough
(c) there is no bounds checking when stashing these stats, so problems
    (a) and (b) can easily lead to buffer overflows

Let's patch this by setting a slightly-more-correct max (that accounts
for a combination of both 2.4G and 5G bands) and adding a bounds check
when writing to our statistics buffer.

Due to problem (b), we still might not properly report all known survey
information (e.g., with "iw <dev> survey dump"), since duplicate results
(or otherwise "larger than expected" results) will cause some
truncation. But that's a problem for a future bugfix.

(And because of this known deficiency, only log the excess at the WARN
level, since that isn't visible by default in this driver and would
otherwise be a bit too noisy.)

Fixes: bf35443314ac ("mwifiex: channel statistics support for mwifiex")
Cc: <stable@vger.kernel.org>
Cc: Avinash Patil <patila@marvell.com>
Cc: Xinming Hu <huxm@marvell.com>
Signed-off-by: Brian Norris <briannorris@chromium.org>
Reviewed-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Ganapathi Bhat <gbhat@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

Merge branch 'qed-next'

Manish Chopra says:

====================
qed/qede: Enhancements

This patch series adds these below features support in qed/qede

1) Ntuple filter configuration [via ethtool -n/N]
2) EEE (energy efficient ethernet) support [ethtool --set-eee/show-eee]
3) Coalescing configuration support for VFs [via ethtool -c/C]

Please consider applying this to "net-next"

V1->V2:
* Fixes below Kbuild test robot warning.

drivers/net//ethernet/qlogic/qed/qed_l2.c:
In function 'qed_get_queue_coalesce':
drivers/net//ethernet/qlogic/qed/qed_l2.c:2137:8: error:
implicit declaration of function 'qed_vf_pf_get_coalesce'
[-Werror=implicit-function-declaration]
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

qed: enhanced per queue max coalesce value.

Maximum coalesce per Rx/Tx queue is extended from
255 to 511.

Signed-off-by: Rahul Verma <rahul.verma@cavium.com>
Signed-off-by: Yuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Read per queue coalesce from hardware

Retrieve the actual coalesce value from hardware for every Rx/Tx
queue, instead of Rx/Tx coalesce value cached during set coalesce.

Signed-off-by: Rahul Verma <Rahul.Verma@cavium.com>
Signed-off-by: Yuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Add support for vf coalesce configuration.

This patch add the ethtool support to set RX/Tx coalesce
value to the VF associated Rx/Tx queues.

Signed-off-by: Rahul Verma <Rahul.Verma@cavium.com>
Signed-off-by: Yuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qede: Add ethtool support for Energy efficient ethernet.

The patch adds ethtool callback implementations for querying/configuring
the Energy Efficient Ethernet (EEE) parameters.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Yuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Add support for Energy efficient ethernet.

The patch adds required driver support for reading/configuring the
Energy Efficient Ethernet (EEE) parameters.

Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@cavium.com>
Signed-off-by: Yuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed/qede: Add setter APIs support for RX flow classification

This patch adds support for adding and deleting rx flow
classification rules. Using this user can classify RX flow
constituting of TCP/UDP 4-tuples [src_ip/dst_ip and src_port/dst_port]
to be steered on a given RX queue

Signed-off-by: Manish Chopra <manish.chopra@cavium.com>
Signed-off-by: Yuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qede: Add getter APIs support for RX flow classification

This patch adds support for ethtool getter APIs to query
RX flow classification rules.

Signed-off-by: Manish Chopra <manish.chopra@cavium.com>
Signed-off-by: Yuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch '40GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
40GbE Intel Wired LAN Driver Updates 2017-07-25

This series contains updates to i40e and i40evf only.

Gustavo Silva fixes a variable assignment, where the incorrect variable
was being used to store the error parameter.

Carolyn provides a fix for a problem found in systems when entering S4
state, by ensuring that the misc vector's IRQ is disabled as well.

Jake removes the single-threaded restriction on the module workqueue,
which was causing issues with events such as CORER.  Does some future
proofing, by changing how the driver displays the UDP tunnel type.

Paul adds a retry in releasing resources if the admin queue times out
during the first attempt to release the resources.

Jesse fixes up references to 32bit timspec, since there are a small set
of errors on 32 bit, so we need to be using the right calls for dealing
with timespec64 variables.  Cleaned up code indentation and corrected
an "if" conditional check, as well as making the code flow more clear.
Cast or changed the types to remove warnings for comparing signed and
unsigned types.  Adds missing includes in i40evf, which were being used
but were not being directly included.

Daniel Borkmann fixes i40e to fill the XDP prog_id with the id just like
other XDP enabled drivers, so that on dump we can retrieve the attached
program based on the id and dump BPF insns, opcodes, etc back to user
space.

Tushar Dave adds le32_to_cpu while evaluating the hardware descriptor
fields, since they are in little-endian format.  Also removed
unnecessary "__packed" to a couple of i40evf structures.

Stefan Assmann fixes an issue when an administratively set MAC was set
and should now be switched back to 00:00:00:00:00:00, the pf_set_mac
flag is not being toggled back to false.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch '10GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
10GbE Intel Wired LAN Driver Updates 2017-07-25

This series contains updates to ixgbe only.

Tony provides all of the changes in the series, starting with adding a
check to ensure that adding a MAC filter was successful, before setting the
MACVLAN.  In order to receive notifications of link configurations of the
external PHY and support the configuration of the internal iXFI link on
X552 devices, Tony enables LASI interrupts.  Update the iXFI driver code
flow, since the MAC register NW_MNG_IF_SEL fields have been redefined for
X553 devices, so add MAC checks for iXFI flows.  Added additional checks
for flow control autonegotiation, since it is not support for X553 fiber
 and XFI devices.

v2: removed unnecessary parens noticed by David Miller in patch 6 of the
    series.
v3: dropped patch 6 of the original series, while we work out a more
    generic solution for malicious driver detection (MDD) support.
v4: updated patch 1 of the series with the comments from Joe Perches which
    were:
      - switched logic to return on error
      - return 0 on success
      - declare retval as an integer
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

bpf: install libbpf headers on 'make install'

Add a new target to install the bpf.h header to $(prefix)/include/bpf/
directory.  This is necessary to build standalone applications using
libbpf, without the need to clone the kernel sources and point to them.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

hamradio: dmascc: avoid -Wformat-overflow warning

gcc warns that the device name might overflow:

drivers/net/hamradio/dmascc.c: In function 'dmascc_init':
drivers/net/hamradio/dmascc.c:584:22: error: 'sprintf' may write a terminating nul past the end of the destination [-Werror=format-overflow=]
   sprintf(dev->name, "dmascc%i", 2 * n + i);
drivers/net/hamradio/dmascc.c:584:3: note: 'sprintf' output between 8 and 17 bytes into a destination of size 16
   sprintf(dev->name, "dmascc%i", 2 * n + i);

>From the static data in this file, I can tell that the index is
strictly limited to 16, so it won't overflow. This simply changes
the sprintf() to snprintf(), which is a good idea in general,
and shuts up this warning.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

i40e: handle setting administratively set MAC address back to zero

When an administratively set MAC was previously set and should now be
switched back to 00:00:00:00:00:00 the pf_set_mac flag did not get
toggled back to false.
As a result VFs were still treated as if an administratively set MAC was
present.

Signed-off-by: Stefan Assmann <sassmann@kpanic.de>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40evf: remove unnecessary __packed

This is similar to 'commit 9588397d24eec ("i40e: remove unnecessary
__packed")' to avoid unaligned access.

Signed-off-by: Tushar Dave <tushar.n.dave@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40evf: Use le32_to_cpu before evaluating HW desc fields

i40e hardware descriptor fields are in little-endian format. Driver
must use le32_to_cpu while evaluating these fields otherwise on
big-endian arch we end up evaluating incorrect values, cause errors
like:
i40evf 0000:03:0a.0: Expected response 24 from PF, received 402653184
i40evf 0000:03:0a.1: Expected response 7 from PF, received 117440512

Signed-off-by: Tushar Dave <tushar.n.dave@oracle.com>
Reviewed-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: report BPF prog id during XDP_QUERY_PROG

Fill the XDP prog_id with the id just like we do in other XDP enabled
drivers such as ixgbe. This is needed so that on dump we can retrieve
the attached program based on the id, and dump BPF insns, opcodes, etc
back to user space. Only XDP driver missing this is currently i40e.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40evf: add some missing includes

These includes were all being used in the driver, but weren't
being directly included.

Since the current advised method is to directly include anything
that you need, this implements that.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: display correct UDP tunnel type name

The i40e driver attempts to display the UDP tunnel name by doing a check
against the type, where for non-zero types we use "vxlan" and for zero
type we use "geneve". This is not future proof, because if new tunnel
types get added, we'll incorrectly label them. It also depends on the
value of UDP_TUNNEL_TYPE_GENEVE == 0, which is brittle.

Instead, replace this with a function that can return a constant string
depending on the type. For now we'll use "unknown" for types we don't
know about, and we can expand this in the future if new types get added.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e/i40evf: remove mismatched type warnings

Compiler reported several places where driver compared
signed and unsigned types.  Cast or change the types to remove
the warnings.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e/i40evf: make IPv6 ATR code clearer

This just reorders some local vars and makes the code flow
clearer.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: fix odd formatting and indent

The compiler warned on an oddly indented bit of code, and when
investigating that, noted that the functions themselves had
an odd flow. The if condition was checked, and would exclude
a call to AQ, but then the aq_ret would be checked unconditionally
which just looks really weird, and is likely to cause objections.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: fix up 32 bit timespec references

As it turns out there was only a small set of errors
on 32 bit, and we just needed to be using the right calls
for dealing with timespec64 variables.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: Handle admin Q timeout when releasing NVM

There are some rare cases where the release resource call will return an
admin Q timeout. In these cases the code needs to try to release the
resource again until it succeeds or it times out.

Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: remove WQ_UNBOUND and the task limit of our workqueue

During certain events such as a CORER, multiple devices will run a work
task to handle some cleanup. This can cause issues due to
a single-threaded workqueue which can mean that a device doesn't cleanup
in time. Prevent this by removing the single-threaded restriction on the
module workqueue. This avoids the need to add more complex yielding
logic in our service task routine. This is also similar to what other
drivers such as fm10k do.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: Fix for trace found with S4 state

This patch fixes a problem found in systems when entering
S4 state.  This patch fixes the problem by ensuring that
the misc vector's IRQ is disabled as well.  Without this
patch a stack trace can be seen upon entering S4 state.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: fix incorrect variable assignment

Fix incorrect variable assignment.
Based on line 1511: aq_ret = I40_ERR_PARAM; the correct variable to be
used in this instance is aq_ret instead of ret. Also, variable ret is
updated at line 1602 just before return, so assigning a value to this
variable in this code block is useless.

Addresses-Coverity-ID: 1397693
Signed-off-by: Gustavo A R Silva <garsilva@embeddedor.com>
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

virtio-net: mark PM functions as __maybe_unused

After removing the reset function, the freeze and restore functions
are now unused when CONFIG_PM_SLEEP is disabled:

drivers/net/virtio_net.c:1881:12: error: 'virtnet_restore_up' defined but not used [-Werror=unused-function]
 static int virtnet_restore_up(struct virtio_device *vdev)
drivers/net/virtio_net.c:1859:13: error: 'virtnet_freeze_down' defined but not used [-Werror=unused-function]
 static void virtnet_freeze_down(struct virtio_device *vdev)

A more robust way to do this is to remove the #ifdef around the callers
and instead mark them as __maybe_unused. The compiler will now just
silently drop the unused code.

Fixes: 4941d472bf95 ("virtio-net: do not reset during XDP set")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: Disable flow control for XFI

Flow control autonegotiation is not supported for XFI.  Make sure that
ixgbe_device_supports_autoneg_fc() returns false and
hw->fc.disable_fc_autoneg is set to true to avoid running the fc_autoneg
function for that device.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: Do not support flow control autonegotiation for X553

Flow control autonegotiation is not supported for fiber on X553.  Add
device ID checks in ixgbe_device_supports_autoneg_fc() to return the
appropriate value.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: Update NW_MNG_IF_SEL support for X553

The MAC register NW_MNG_IF_SEL fields have been redefined for
X553. These changes impact the iXFI driver code flow. Since iXFI is
only supported in X552, add MAC checks for iXFI flows.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: Enable LASI interrupts for X552 devices

Enable LASI interrupts on X552 devices in order to receive notifications of
link configurations of the external PHY and support the configuration of
the internal iXFI link since iXFI does not support auto-negotiation.  This
is not required for X553 devices; add a check to avoid enabling LASI
interrupts for X553 devices.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: Ensure MAC filter was added before setting MACVLAN

This patch adds a check to ensure that adding the MAC filter was
successful before setting the MACVLAN.  If it was unsuccessful, propagate
the error.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>