The debug code dereferences "skb" to print "skb->len" so we have to
print the message before we free "skb".
Fixes: f99fe49ff3 ("wil6210: add wil_netif_rx() helper function")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Fix a race between cfg80211 add_key call and transmitting of 4/4 EAP
packet. In case the transmit is delayed until after the add key takes
place, message 4/4 will be encrypted with the new key, and the
receiver side (AP) will drop it due to MIC error.
Wil6210 will monitor and look for the transmitted packet 4/4 eap key.
In case add_key takes place before the transmission completed, then
wil6210 will let the FW store the key and wil6210 will notify the FW
to use the PTK key only after 4/4 eap packet transmission was
completed.
Signed-off-by: Ahmad Masri <amasri@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Move common part of wil_netif_rx_any into new helper function and add
support for non-gro receive using netif_rx_ni.
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
In preparation for unifying the skb_frag and bio_vec, use the fine
accessors which already exist and use skb_frag_t instead of
struct skb_frag_struct.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace all wmi_call timeouts that are less than 100 msec to use
WIL_WMI_CALL_GENERAL_TO_MS (100 msec) as a default. Some of the
current wmi_call timeouts are too short and fails to receive its
waiting events.
Signed-off-by: Ahmad Masri <amasri@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Save the parameter of max_assoc_sta per device.
On fw_ready event calculate the max_assoc_sta and save it per device,
do not overwrite the max_assoc_sta module param.
Signed-off-by: Ahmad Masri <amasri@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
_desc is defined in __wil_tx_vring_tso() and may not be set in
case len is 0, verify _desc is set.
Signed-off-by: Alexei Avshalom Lazar <ailizaro@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Notice that *rc* can evaluate to up to 5, include/linux/netdevice.h:
enum gro_result {
GRO_MERGED,
GRO_MERGED_FREE,
GRO_HELD,
GRO_NORMAL,
GRO_DROP,
GRO_CONSUMED,
};
typedef enum gro_result gro_result_t;
In case *rc* evaluates to 5, we end up having an out-of-bounds read
at drivers/net/wireless/ath/wil6210/txrx.c:821:
wil_dbg_txrx(wil, "Rx complete %d bytes => %s\n",
len, gro_res_str[rc]);
Fix this by adding element "GRO_CONSUMED" to array gro_res_str.
Addresses-Coverity-ID: 1444666 ("Out-of-bounds read")
Fixes: 194b482b50 ("wil6210: Debug print GRO Rx result")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Reviewed-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Rearrange the code by having functions to access 802.3 header
members, source and destination addresses.
Signed-off-by: Ahmad Masri <amasri@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
New FW added support for upto 20 clients in AP mode. Change the driver
to support this as well. FW reports it's max supported associations in
WMI_READY_EVENT. Some WMI commands/events use cidxtid field which is
limited to 16 cids. Use new cid/tid fields instead.
For Rx packets cid from rx descriptor is limited to 3 bits (0..7),
to find the real cid, compare transmitter address with the stored
stations mac address in the driver sta array.
EDMA FW still supports 8 stations. Extending the support to 20
stations will come later.
Signed-off-by: Ahmad Masri <amasri@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
In AP mode with multiple clients, driver stops net queue
(netif_tx_stop_queue) upon first ring (serving specific client)
becoming full. This can have negative effect on transmission to
other clients which may still have room in their corresponding rings.
Implement new policy in which stop/wake net queue are not used. In
case there is no room in the ring for a transmitted packet, drop the
packet.
New policy can be helpful to debug performance issues, to guarantee
maximum utilization of net queues.
New policy is disabled by default and can be enabled by debugfs:
echo 1 > drop_if_ring_full
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Due to a HW issue in PHY info collection rtap_include_phy_info is not
in use, hence can be removed.
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
A successful call to wil_tx_ring takes skb reference so
it will only be freed in wil_tx_complete. Consume the skb
in wil_find_tx_bcast_2 to prevent memory leak.
Signed-off-by: Lior David <liord@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
When enlarging rx_ring_order module param, wil6210 fails to load
because there are not enough Rx buffers.
Fix this by enlarging number of Rx buffers at startup, if needed based
on rx_ring_order.
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Setting the same fake hash to all skbs prevents
distributing different flows to different CPU cores.
Signed-off-by: Hamad Kadmany <hkadmany@codeaurora.org>
Signed-off-by: Lior David <liord@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
This feature is needed for enterprise APs and clients to enable
fast roaming as defined in 802.11r between APs in the same ESS.
On AP side, this feature is supported only when disable_ap_sme
is enabled.
Signed-off-by: Ahmad Masri <amasri@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
In enhanced DMA ring 0 is used for RX ring, hence TX ring 0
is an unused element in ring_tx and ring2cid_tid arrays.
Initialize ring2cid_tid CID to WIL6210_MAX_CID to prevent a false
match of CID 0.
Go over the ring_tx and ring2cid_tid from wil_get_min_tx_ring_id
and on to prevent access to index 0 in eDMA.
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Delivering a looped-back multicast packet to network stack can cause
higher layer protocols to fail like for example IPv6 DAD.
In STA mode, upon receiving Rx multicast packet, check if the source
MAC address is equal to our own MAC address and if so drop the packet.
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Currently the driver sets CHECKSUM_UNNECESSARY only in case the HW
doesn't report checksum error.
As ip_summed value is not initialized it is not clear what
the driver will report to the network stack in case of HW checksum
error or in case HW doesn't calculate checksum.
Initialize ip_summed to CHECKSUM_NONE to guarantee checksum
calculation by the network stack in the above cases.
Signed-off-by: Gidon Studinski <gidons@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Collect statistics of TX latency. The latency is measured from the time
the HW gets aware of new SKB to transmit until the HW indicates tx
complete for this SKB.
The statistics are shown via new "tx_latency" debugfs.
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Due to recent change in FW, driver will be notified of corrupted Rx
packets (e.g. MIC error).
Drop such packets before they are delivered to network stack.
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Store the last received multicast sequence number (SN) part of the TID
info. Drop Rx multicast packets with retry bit set which their SN
is equal to the last received.
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Enhanced DMA RX data path is handled using a single
RX descriptor ring for all VIFs.
Multiple RX status rings are supported, to allow RSS
and multi MSI support.
The driver gets the RX completions via the RX status rings.
The RX status message includes the completed RX buffer ID,
which points to the allocated SKB.
The enhanced DMA RX data flow supports RX chaining, where
multiple SKBs are merged into a single packet.
Enhanced DMA HW supports RX HW reorder offload, enabled by
default for Talyn-MB.
amsdu_en debugfs entry was added to allow control MSDU aggregation.
Use the following command to disable AMSDU (enabled by default):
echo 0 > amsdu_en
Signed-off-by: Gidon Studinski <gidons@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The enhanced DMA TX data path is handled using a descriptor
ring per connection and a single status ring.
The driver gets TX completions via the TX status ring. Each
status message points to the completed descriptor ring and
includes the number of completed descriptors in this ring.
Non TSO enhanced DMA TX descriptors are similar to legacy DMA
TX descriptors, hence the same transmit function can be used.
However, enhanced DMA TSO frames division is performed by the
HW, hence a new function is added to handle enhanced DMA TSO.
Signed-off-by: Gidon Studinski <gidons@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Enhanced DMA design includes the following rings:
- Single RX descriptor ring is used for all VIFs
- Multiple RX status rings are supported, to allow RSS
- TX descriptor ring is allocated per connection
- A single TX status ring is used for all TX descriptor rings
This patch initializes and frees the above descriptor and
status rings.
The RX SKBs are handled by a new entity of RX buffers manager,
which handles RX buffers, each one points to an allocated SKB.
During Rx completion processing, the driver extracts a buffer
ID which is used as an index to the buffers array.
After the SKB is freed the buffer is moved from the 'active'
list to the 'free' list, indicating it can be used for another
descriptor. During Rx refill, SKBs are allocated and attached
to 'free' buffers. Those buffers are attached to new descriptors
and moved to the 'active' list.
New debugfs entries were added to allow edma configuration:
Run the following command to configure the number of status rings:
echo NUM_OF_STATUS_RINGS > num_rx_status_rings
Run the following command to use extended RX status message for
additional debug fields from HW:
echo 0 > compressed_rx_status
Run the following command to control the size of the TX status ring:
echo TX_STATUS_RING_ORDER > tx_status_ring_order
The status ring size will be 1 << tx_status_ring_order
Run the following command to control the size of the RX status ring:
echo RX_STATUS_RING_ORDER > rx_status_ring_order
Due to HW constrains RX sring order should be bigger than RX ring order
The status ring size will be 1 << rx_status_ring_order
Run the following command to change the number of RX buffer IDs:
echo RX_BUFF_ID_COUNT > rx_buff_id_count
Signed-off-by: Gidon Studinski <gidons@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
In enhanced DMA the vrings are handled internally by the FW
and are not exposed to the driver.
Instead, the driver handles descriptor rings, which are mapped
by the FW to vrings.
The completions of the TX and RX descriptors are notified to
the driver using status rings. Each status ring descriptor
includes information of the completed descriptors and the ring id
of their descriptor ring.
This patch changes struct vring to generic wil_ring to allow
its reuse for enhanced DMA descriptor rings and adds the descriptor
and status rings specific descriptors.
The vring debugfs entries have changed as follows:
- dbg_vring_index has changed to dbg_ring_index
- vrings has changed to rings
- vring_idle_trsh has changed to ring_idle_trsh
- vring_index has changed to ring_index
Signed-off-by: Gidon Studinski <gidons@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
wil_err inside wil_rx_refill can flood the log buffer.
Replace it with wil_err_ratelimited.
Signed-off-by: Dedy Lansky <dlansky@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
WMI command reply saved in uninitialized struct.
In order to avoid accessing unset values from FW initialize
the reply struct.
Signed-off-by: Alexei Avshalom Lazar <ailizaro@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Track the connection status per-VIF.
The data path code is also updated to support multiple VIFs.
This includes RX and TX VRING management, NAPI poll loops,
RX reordering and related code.
Power management code used to check if the main interface
is up or based on connection state of the main interface,
adapt this code to take all VIFs into account.
Signed-off-by: Lior David <liord@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Simple infrastructure changes for supporting multiple
virtual interfaces (multiple VIFs).
It is still not possible to add new VIFs so the only VIF
belongs to the main interface.
Main changes:
1. Add MAC ID(mid) argument to wmi_send and wmi_call to
allow invoking WMI commands on different VIFs.
2. Similarly, in WMI event handler look at the mid reported
by FW and extract VIF structure (currently only for main
interface). All WMI event handlers operate on wil6210_vif
structure so they know on which VIF they were called.
3. Trivial changes to use wil6210_vif structure and MID
throughout the code.
4. Various changes to logging to report MID.
More complete multiple VIFs support will be added gradually
in next patches.
Signed-off-by: Lior David <liord@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The field preset_chandef of wireless_dev must not be accessed
by the driver because it is private to cfg80211. Store the
monitor channel locally in wil6210_priv instead.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Add the option to support 40bit addresses since some platforms
may not support 48bits but support 40bits
Signed-off-by: Lazar Alexei <qca_ailizaro@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
vring_idle_trsh is used in the operational driver, hence
should not be defined as a debugfs variable.
Signed-off-by: Gidon Studinski <qca_gidons@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
New features and bug fixes to quite a few different drivers, but
nothing really special standing out.
What makes me happy that we have now more vendors actively
contributing to upstream drivers. In this pull request we have patches
from Broadcom, Intel, Qualcomm, Realtek and Redpine Signals, and I
still have patches from Marvell and Quantenna pending in patchwork. Now
that's something comparing to how things looked 11 years ago in Jeff
Garzik's "State of the Union: Wireless" email:
https://lkml.org/lkml/2006/1/5/671
Major changes:
wil6210
* add low level RF sector interface via nl80211 vendor commands
* add module parameter ftm_mode to load separate firmware for factory
testing
* support devices with different PCIe bar size
* add support for PCIe D3hot in system suspend
* remove ioctl interface which should not be in a wireless driver
ath10k
* go back to using dma_alloc_coherent() for firmware scratch memory
* add per chain RSSI reporting
brcmfmac
* add support multi-scheduled scan
* add scheduled scan support for specified BSSIDs
* add support for brcm43430 revision 0
wlcore
* add wil1285 compatible
rsi
* add RS9113 USB support
iwlwifi
* FW API documentation improvements (for tools and htmldoc)
* continuing work for the new A000 family
* bump the maximum supported FW API to 31
* improve the differentiation between 8000, 9000 and A000 families
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJZT/ROAAoJEG4XJFUm622b/IwH+wQtBXP+P57DZ97BdETcZ3Wd
ejm9KKT3IxcP7kzItT9UR1zv7LNx4NgSkJMfhf37jJ1WIuE2fp/ctNv6mz3PvohW
jD1fLynwEMMC7PLHEy5+xGtL61KYc2mtXs/bfLFl94hZUiaocrsBDIT4fXoyIWif
y3MUlBKDbHA27ULRd485C0MRekRSvR/rq6iST4KsIsa8RflJbdH64teTEnQPp4kh
nvhfnaxVEqHK7mYbarC58yYLNOU8gfQXmeeTfd6jhCXRjfEw37IvtNC8BT4B9ZDX
YMZLbkeDSwNGXIH7EIrpGOYHCPh+qsiJ+sx7YodomcJvd1nVrmbuZZt9/EqWXuY=
=Fd/c
-----END PGP SIGNATURE-----
Merge tag 'wireless-drivers-next-for-davem-2017-06-25' of git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers-next
Kalle Valo says:
====================
wireless-drivers-next patches for 4.13
New features and bug fixes to quite a few different drivers, but
nothing really special standing out.
What makes me happy that we have now more vendors actively
contributing to upstream drivers. In this pull request we have patches
from Broadcom, Intel, Qualcomm, Realtek and Redpine Signals, and I
still have patches from Marvell and Quantenna pending in patchwork. Now
that's something comparing to how things looked 11 years ago in Jeff
Garzik's "State of the Union: Wireless" email:
https://lkml.org/lkml/2006/1/5/671
Major changes:
wil6210
* add low level RF sector interface via nl80211 vendor commands
* add module parameter ftm_mode to load separate firmware for factory
testing
* support devices with different PCIe bar size
* add support for PCIe D3hot in system suspend
* remove ioctl interface which should not be in a wireless driver
ath10k
* go back to using dma_alloc_coherent() for firmware scratch memory
* add per chain RSSI reporting
brcmfmac
* add support multi-scheduled scan
* add scheduled scan support for specified BSSIDs
* add support for brcm43430 revision 0
wlcore
* add wil1285 compatible
rsi
* add RS9113 USB support
iwlwifi
* FW API documentation improvements (for tools and htmldoc)
* continuing work for the new A000 family
* bump the maximum supported FW API to 31
* improve the differentiation between 8000, 9000 and A000 families
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to preserve the connection in suspend/resume flow,
wil6210 host allows going to PCIe D3hot state in suspend,
instead of performing a full wil6210 device reset. This
requires the platform ability to initiate wakeup in case of
RX data. To check that, a new platform API is added.
In addition, add cfg80211 suspend/resume callbacks
implementation.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
It seems like a historic accident that these return unsigned char *,
and in many places that means casts are required, more often than not.
Make these functions return void * and remove all the casts across
the tree, adding a (u8 *) cast only where the unsigned char pointer
was used directly, all done with the following spatch:
@@
expression SKB, LEN;
typedef u8;
identifier fn = { skb_push, __skb_push, skb_push_rcsum };
@@
- *(fn(SKB, LEN))
+ *(u8 *)fn(SKB, LEN)
@@
expression E, SKB, LEN;
identifier fn = { skb_push, __skb_push, skb_push_rcsum };
type T;
@@
- E = ((T *)(fn(SKB, LEN)))
+ E = fn(SKB, LEN)
@@
expression SKB, LEN;
identifier fn = { skb_push, __skb_push, skb_push_rcsum };
@@
- fn(SKB, LEN)[0]
+ *(u8 *)fn(SKB, LEN)
Note that the last part there converts from push(...)[0] to the
more idiomatic *(u8 *)push(...).
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The 11ad spec requires 11ad devices to be able to receive 8KB
packets over the air. Currently this is only possible by
loading the driver with mtu_max=7912 but this also forces
a smaller block ACK window size which reduces performance
for stations which transmit normal sized packets (<2KB).
Fix this problem as follows:
1. Add a module parameter rx_large_buf that when set,
will allocate 8KB RX buffers regardless of mtu_max
setting.
2. When receiving block ACK request agree to any window
size not above our maximum, regardless of the mtu_max setting.
This means if the other side transmits small packets (2KB)
it can still set up block ACK with a large window size,
and get better performance.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
device supports 48bit addresses, reflect that by
setting the dma mask accordingly.
Signed-off-by: Hamad Kadmany <qca_hkadmany@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Symbolic permissions are no longer recommended.
This patch changes the symbolic permissions in wil6210 driver
to octal permissions.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Currently we do not have full support for broadcast from
a station inside a PBSS network.
We have a workaround where instead of broadcast we do a
unicast to every known station in the PBSS.
This workaround was performed only for P2P clients.
This fix will perform the broadcast workaround also for a
regular station inside a PBSS.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
__func__ is automatically added to printouts by dynamic debug
mechanism and by wil_info/wil_err macros.
Remove __func__ from debug printouts to avoid duplication.
Signed-off-by: Lazar Alexei <qca_ailizaro@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Driver calls to netif_tx_stop_all_queues/netif_tx_wake_all_queues are
inconsistent. In several cases, driver can get to a situation where net
queues are stopped forever and data cannot be sent.
The fix is to stop net queues if there is at least one vring which is
"full" and to wake net queues if all vrings are not "full".
Signed-off-by: Dedy Lansky <qca_dlansky@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
txdata->enabled is used in order to determine if the TX vring
is valid. As the data transmit is handled in a different context,
in case txdata->enabled is set before vring->hwtail is updated,
an old or corrupted vring->hwtail can be used.
Protect setting of txdata->enabled and vring->hwtail to prevent a
case where TX vring start handling TX packets before setting
vring->hwtail.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
In case we fail to map one of the TSO SKB fragments, we need to
clear all the mapped descriptors, from swhead to swhead+descs_used-1.
Change the desc index calculation to
i = (swhead + descs_used - 1) % vring->size;
to prevent unmpping of (swhead + descs_used) descriptor that wasn't
mapped.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
napi_synchronize is called before releasing the vring, with the
assumption that setting txdata->enabled to 0 will prevent handling
of this vring in the next scheduled napi.
To guarantee this assumption, a memory barrier is added after disabling
the txdata.
In addition, as the ctx is zeroed in wil_tx_complete after this
descriptor is handled (protected by wmb), ctx needs to be checked
before releasing this descriptor in wil_vring_free.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
add memory barrier after allocating new rx descriptors, before
updating the hwtail.
This will guarantee that all writes to descriptors (shared memory)
are done before committing them to HW.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
There are 2 possible race conditions, both are solved by addition of
memory barrier:
1. wil_tx_complete reads the swhead to determine if the vring is
empty. In case the swhead was updated before the descriptor update
was performed in __wil_tx_vring/__wil_tx_vring_tso, the completion
loop will not end and as the DU bit may still be set from a previous
run, this skb can be handled as completed before it was sent, which
will lead to double free of the same SKB.
2. __wil_tx_vring/__wil_tx_vring_tso calculate the number of available
descriptors according to the swtail. In case the swtail is updated
before memset of ctx to zero is completed, we can handle this
descriptor while later on ctx is zeroed.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Network stack can try to transmit data while AP / STA is
disconnected.
Change this print-out to debug level as this should not be
handled as error.
This patch also adds wil_dbg_ratelimited, used to limit the
above print-out.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Dan Carpenter
Ahmad Masri
Dedy Lansky
Matthew Wilcox (Oracle)
Alexei Avshalom Lazar
Gustavo A. R. Silva
Maya Erez
Lior David
Hamad Kadmany
Gidon Studinski
Lior David
Lazar Alexei
Gidon Studinski
David S. Miller
Maya Erez
Johannes Berg
Hamad Kadmany
Linus Torvalds
Masanari Iida
Dedy Lansky