In case a sync command timeouts or Tx is stuck while a FW error
interrupt arrives, we might call iwl_op_mode_nic_error twice before
a restart has been initiated. This will cause a reprobe. Unify calls
to this function at the transport level and only call it on the first
FW error in a given by checking the transport FW error flag.

While at it, remove the privately defined iwl_nic_error from PCIE code
and use the common callback instead.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

The same bits are employed in all transport layers. Put the status
field in the common transport layer. This allows us to employ them
in common transport code.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

If we call ieee80211_hw_restart, it means that the
firmware is in bad condition and will be reset soon.
Since the firmware will be reset, there is no good
reason to keep sending host commands.
Signed-off-by: NAlexander Bondar <alexander.bondar@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>

This means it can be shared for different transport
layers in the future.
Signed-off-by: NInbal Hacohen <Inbal.Hacohen@intel.com>
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>

Accessing the device in Tx path is not a good idea.
Mirror the data in DRAM.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Recently in commit 8a964f44
("iwlwifi: always copy first 16 bytes of commands") we fixed
the problem that the hardware writes back to the command and
that could overwrite parts of the data that was still needed
and would thus be corrupted.

Investigating this problem more closely we found that this
write-back isn't really ordered very well with respect to
other DMA traffic. Therefore, it sometimes happened that the
write-back occurred after unmapping the command again which
is clearly an issue and could corrupt the next allocation
that goes to that spot, or (better) cause IOMMU faults.

To fix this, allocate coherent memory for the first 16 bytes
of each command, containing the write-back part, and use it
for all queues. All the dynamic DMA mappings only need to be
TO_DEVICE then. This ensures that even when the write-back
happens "too late" it can't hit memory that has been freed
or a mapping that doesn't exist any more.

Since now the actual command is no longer modified, we can
also remove CMD_WANT_HCMD and get rid of the DMA sync that
was necessary to update the scratch pointer.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

The FH hardware will always write back to the scratch field
in commands, even host commands not just TX commands, which
can overwrite parts of the command. This is problematic if
the command is re-used (with IWL_HCMD_DFL_NOCOPY) and can
cause calibration issues.

Address this problem by always putting at least the first
16 bytes into the buffer we also use for the command header
and therefore make the DMA engine write back into this.

For commands that are smaller than 16 bytes also always map
enough memory for the DMA engine to write back to.

Cc: stable@vger.kernel.org
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

With new transports coming up, move to threaded
interrupt handling now. This has the advantage
that we can use the same locking scheme with all
different transports we may need to implement.

Note that the TX path obviously still runs in a
tasklet, so some spin_lock() calls need to change
to spin_lock_bh() calls to properly lock out the
TX path.

In my test on a Calpella platform this has no
impact on throughput or latency.

Also add lockdep annotations to avoid lockups due
to catch sending synchronous commands or using
locks that connect with them from the irq thread.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Move the reg_lock that protects HW register access
into the transport implementation. Locking is no
longer exposed, but handled internally in grab and
release NIC access. This simplifies the users.
Signed-off-by: NLilach Edelstein <lilach.edelstein@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Update Copyright notices to 2013.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

On resuming, the opmode may have to be able to talk
to the WoWLAN/D3 firmware in order to query it about
its status and wakeup reasons. To do that, the opmode
has to call the new d3_resume() transport API which
will set up the device for command communcation.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

There's no need to request the IRQ every time the
device is started, we can request it just once.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This is needed for newer NICs.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

It is not used outside pcie/rx.c.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Also, prefer the if(!X) notation over if(X == 0).
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Rename static functions. Function moved from trans.c to
tx.c. A few could be made static, others had to be exported.

Functions that implement the transport API are prefixed by
iwl_trans_pcie_, the others by iwl_pcie_.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Rename static functions. Function moved from trans.c to
rx.c. A few could be made static, others had to be exported.
Also, don't use rxb or rxbuf, but rb which stands for receive
buffer.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

1) s/tx_queue/txq
	for the sake of consistency.

2) s/rx_queue/rxq
	for the sake of consistency.

3) Make all functions begin with iwl_pcie_
	iwl_queue_init and iwl_queue_space are an exception
	since they are not PCIE specific although they are in
	pcie subdir.

4) s/trans_pcie_get_cmd_string/get_cmd_string
	it is much shorter and used in debug prints which
	are long lines.

5) s/iwl_bg_rx_replenish/iwl_pcie_rx_replenish_work
	this better emphasizes that it is a work

6) remove invalid kernelDOC markers

pcie/tx.c and pcie/trans.c still needs to be cleaned up.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

When a firmware error occurs, don't just abort synchronous
commands but also return an error (-EIO) and block any new
commands as well. Currently, an error is only returned if
WANT_SKB was set which is confusing and can lead to issues.

Blocking is done until a new firmware image is loaded.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

No HCMD can be sent while RFKILL is asserted. If a SYNC
command is running while RFKILL is asserted the fw will
silently discard it. This means that the driver needs to
wake the process that sleeps on the CMD_SYNC.

Since the RFKILL interrupt is handled in the transport layer
and the code that sleeps in CMD_SYNC is also in the transport
layer, all this logic can be handled there.
This simplifies the work of the op_mode.

So the transport layer will now return -ERFKILL when a CMD
is sent and RFKILL is asserted. This will be the case even
when the CMD is SYNC. The transport layer will return
-ERFKILL straight away.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

In addition to the NOCOPY flag, add a DUP flag that
tells the transport to kmemdup() the buffer and free
it after the command completes.

Currently this is only supported for a single buffer
in a given command, but that could be extended if it
should be needed.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Since the queue might not be empty, we need to free the
pending Tx packets.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Remove outdated iwlagn prefix to a few functions and fix comments
that were not accurate.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Reviewed-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

iwl_dbgfs_fh_reg_read() can cause crashes and/or
BUG_ON in slub because the ifdefs are wrong, the
code in iwl_dump_fh() should use DEBUGFS, not
DEBUG to protect the buffer writing code.

Also, while at it, clean up the arguments to the
function, some code and make it generally safer.

Cc: stable@vger.kernel.org
Reported-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Until now, the response handler of a Host Command got the
exact same pointer that was also given to the DMA engine.
We almost never need to the Host Command that was sent while
handling its response, but when we do need it, we see that
the command has been modified.

This mystery has been elucidated. The FH (our DMA engine)
writes its meta data on the buffer in the DRAM. Of course it
copies the buffer to the NIC first. This was known to happen
for Tx command, but as a matter of fact, it happens to all
TFD brought by the FH which doesn't care much about what it
brings from DRAM to internal SRAM.

So copy the Host Command to yet another buffer so that we
can properly pass the buffer that was sent originally to the
fw. Do that only if it was request by the user since very
few flows need to get the HCMD sent in the response handler.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Since the op_mode defines the queue mapping, let it do it
completely through the API functions.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This locking isn't needed. The only locking we need is when
we access prph registers but there is already a separate
lock for that.
Since we haven't returned from the mac80211's
IEEE80211_AMPDU_TX_OPERATIONAL ampdu_action, we cannot
receive any Tx frame for that sta / tid while enabling the
queue.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This is not needed, we just need to tell the SCD not to use
that queue. We will reconfigure that queue when we will use
it again.

Clean up a bit the code on the way.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

We need to be able to enable / disable Tx queues in HW
dynamically. So this function is no longer related to AGG
only. It can do the job for any queue, even AC ones. Change
the name to better reflect its role.

Also use the new function to configure the AC / CMD queues
in tx_start.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

We need to be able to enable / disable Tx queues in HW
dynamically. So this function is no longer related to AGG
only. It can do the job for any queue, even AC ones. Change
the name to better reflect its role.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Structure the code a bit more and move all PCIe code
including the hardware configuration files into a
PCIe specific subdirectory.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Mostly clean up indentation around parentheses
after if, function calls, etc. and also a few
unneeded line breaks and some other things.
Reviewed-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Reviewed-by: NWey-Yi W Guy <wey-yi.w.guy@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Reviewed-by: NWey-Yi W Guy <wey-yi.w.guy@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The logic that allows to have a short TFD queue was completely wrong.
We do maintain 256 Transmit Frame Descriptors, but they point to
recycled buffers. We used to attach and de-attach different TFDs for
the same buffer and it worked since they pointed to the same buffer.

Also zero the number of BDs after unmapping a TFD. This seems not
necessary since we don't reclaim the same TFD twice, but I like
housekeeping.

This patch solves this warning:

[ 6427.079855] WARNING: at lib/dma-debug.c:866 check_unmap+0x727/0x7a0()
[ 6427.079859] Hardware name: Latitude E6410
[ 6427.079865] iwlwifi 0000:02:00.0: DMA-API: device driver tries to free DMA memory it has not allocated [device address=0x00000000296d393c] [size=8 bytes]
[ 6427.079870] Modules linked in: ...
[ 6427.079950] Pid: 6613, comm: ifconfig Tainted: G           O 3.3.3 #5
[ 6427.079954] Call Trace:
[ 6427.079963]  [<c10337a2>] warn_slowpath_common+0x72/0xa0
[ 6427.079982]  [<c1033873>] warn_slowpath_fmt+0x33/0x40
[ 6427.079988]  [<c12dcb77>] check_unmap+0x727/0x7a0
[ 6427.079995]  [<c12dcdaa>] debug_dma_unmap_page+0x5a/0x80
[ 6427.080024]  [<fe2312ac>] iwlagn_unmap_tfd+0x12c/0x180 [iwlwifi]
[ 6427.080048]  [<fe231349>] iwlagn_txq_free_tfd+0x49/0xb0 [iwlwifi]
[ 6427.080071]  [<fe228e37>] iwl_tx_queue_unmap+0x67/0x90 [iwlwifi]
[ 6427.080095]  [<fe22d221>] iwl_trans_pcie_stop_device+0x341/0x7b0 [iwlwifi]
[ 6427.080113]  [<fe204b0e>] iwl_down+0x17e/0x260 [iwlwifi]
[ 6427.080132]  [<fe20efec>] iwlagn_mac_stop+0x6c/0xf0 [iwlwifi]
[ 6427.080168]  [<fd8480ce>] ieee80211_stop_device+0x5e/0x190 [mac80211]
[ 6427.080198]  [<fd833208>] ieee80211_do_stop+0x288/0x620 [mac80211]
[ 6427.080243]  [<fd8335b7>] ieee80211_stop+0x17/0x20 [mac80211]
[ 6427.080250]  [<c148dac1>] __dev_close_many+0x81/0xd0
[ 6427.080270]  [<c148db3d>] __dev_close+0x2d/0x50
[ 6427.080276]  [<c148d152>] __dev_change_flags+0x82/0x150
[ 6427.080282]  [<c148e3e3>] dev_change_flags+0x23/0x60
[ 6427.080289]  [<c14f6320>] devinet_ioctl+0x6a0/0x770
[ 6427.080296]  [<c14f8705>] inet_ioctl+0x95/0xb0
[ 6427.080304]  [<c147a0f0>] sock_ioctl+0x70/0x270

Cc: stable@vger.kernel.org
Reported-by: NAntonio Quartulli <ordex@autistici.org>
Tested-by: NAntonio Quartulli <ordex@autistici.org>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Reviewed-by: NWey-Yi W Guy <wey-yi.w.guy@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Since the transport allocates and frees itself in
the transport specific code, there's no need for
virtual functions for it. Remove the free method
and call the correct functions directly.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NWey-Yi Guy <wey-yi.w.guy@intel.com>

Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NWey-Yi Guy <wey-yi.w.guy@intel.com>

Having cmd[], meta[] and skbs[] as separate arrays
in the TX queue structure is cache inefficient as
we need the data for a given entry together.

To improve this, create an array with these three
members (allocate meta as part of that struct) so
we have the data we need together located together
improving cache footprint.

The downside is that we need to allocate a lot of
memory in one chunk, about 10KiB (on 64-bit) which
isn't very efficient.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NWey-Yi Guy <wey-yi.w.guy@intel.com>

Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NWey-Yi Guy <wey-yi.w.guy@intel.com>

The driver layer now holds a pointer to the transport,
and shrd->drv is not needed any more, so kill it.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NWey-Yi Guy <wey-yi.w.guy@intel.com>