This patch fixes the race condition when one thread tries to destroy
the memory allocated for rx_roq, while another thread still happen
to access rx_roq.
Such a race condition occurs when i2400m-sdio kernel module gets
unloaded, destroying the memory allocated for rx_roq while rx_roq
is accessed by i2400m_rx_edata(), as explained below:
$thread1 $thread2
$ void i2400m_rx_edata() $
$Access rx_roq[] $
$roq = &i2400m->rx_roq[ro_cin] $
$ i2400m_roq_[reset/queue/update_ws] $
$ $ void i2400m_rx_release();
$ $kfree(rx->roq);
$ $rx->roq = NULL;
$Oops! rx_roq is NULL
This patch fixes the race condition using refcount approach.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
This patch increases the tx_queue_len to 20 so as to
minimize the jitter in the throughput.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
This patch fixes an infinite loop caused by i2400m_tx_fifo_push() due
to a corner case where there is no tail space in the TX FIFO.
Please refer the documentation in the code for details.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
This fixes i2400m_tx_fifo_push(); the check for having enough
space in the TX FIFO's tail was obscure and broken in certain
corner cases. The new check works in all cases and is way
clearer. Please refer the documentation in the code for details.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
The older method of computing the maximum PDU size relied
on a method that doesn't work when we prop the maximum
number of payloads up to the physical limit, and thus we kill
the whole computation and just verify that the constants are
congruent.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
According to Intel Wimax i3200, i5x50 and i6x50 specification
documents, the maximum size of each TX message can be upto 16KiB.
This patch modifies the i2400m_tx() routine to check that the
message size does not exceed the 16KiB limit.
Please refer the documentation in the code for details.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
According to Intel Wimax i3200, i5x50 and i6x50 device specification
documents, the maximum number of payloads per message can be up to 60.
Increasing the number of payloads to 60 per message helps to
accommodate smaller payloads in a single transaction. This patch
increases the maximum number of payloads from 12 to 60 per message.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
This patch makes sure whenever tx_setup() is invoked during driver
initialization or device reset where TX FIFO is released and re-allocated,
the indices tx_in, tx_out, tx_msg_size, tx_sequence, tx_msg are properly
initialized.
When a device reset happens and the TX FIFO is released/re-allocated,
a new block of memory may be allocated for the TX FIFO, therefore tx_msg
should be cleared so that no any TX threads (tx_worker, tx) would access
to the out-of-date addresses.
Also, the TX threads use tx_in and tx_out to decide where to put the new
host-to-device messages and from where to copy them to the device HW FIFO,
these indices have to be cleared so after the TX FIFO is re-allocated during
the reset, the indices both refer to the head of the FIFO, ie. a new start.
The same rational applies to tx_msg_size and tx_sequence.
To protect the indices from being accessed by multiple threads simultaneously,
the lock tx_lock has to be obtained before the initializations and released
afterwards.
Signed-off-by: Cindy H Kao <cindy.h.kao@intel.com>
When bus_setup fails in i2400m_post_reset(), it falls to the error path handler
"error_bus_setup:" which includes unlock the mutext. However, we didn't ever
try to the obtain the lock when running bus_setup.
The patch is to fix the misplaced error path handler "error_bus_setup:".
Signed-off-by: Cindy H Kao <cindy.h.kao@intel.com>
This patch adds an error recovery mechanism on TX path.
The intention is to bring back the device to some known state
whenever TX sees -110 (-ETIMEOUT) on copying the data to the HW FIFO.
The TX failure could mean a device bus stuck or function stuck, so
the current error recovery implementation is to trigger a bus reset
and expect this can bring back the device.
Since the TX work is done in a thread context, there may be a queue of TX works
already that all hit the -ETIMEOUT error condition because the device has
somewhat stuck already. We don't want any consecutive bus resets simply because
multiple TX works in the queue all hit the same device erratum, the flag
"error_recovery" is introduced to denote if we are ready for taking any
error recovery. See @error_recovery doc in i2400m.h.
Signed-off-by: Cindy H Kao <cindy.h.kao@intel.com>
The problem is only seen on SDIO interface since on USB, a bus reset would
really re-probe the driver, but on SDIO interface, a bus reset will not
re-enumerate the SDIO bus, so no driver re-probe is happening. Therefore,
on SDIO interface, the reset event should be still detected and handled by
dev_reset_handle().
Problem description:
Whenever a reboot barker is received during operational mode (i2400m->boot_mode == 0),
dev_reset_handle() is invoked to handle that function reset event.
dev_reset_handle() then sets the flag i2400m->boot_mode to 1 indicating the device is
back to bootmode before proceeding to dev_stop() and dev_start().
If dev_start() returns failure, a bus reset is triggered by dev_reset_handle().
The flag i2400m->boot_mode then remains 1 when the second reboot barker arrives.
However the interrupt service routine i2400ms_rx() instead of invoking dev_reset_handle()
to handle that reset event, it filters out that boot event to bootmode because it sees
the flag i2400m->boot_mode equal to 1.
The fix:
Maintain the flag i2400m->boot_mode within dev_reset_handle() and set the flag
i2400m->boot_mode to 1 when entering dev_reset_handle(). It remains 1
until the dev_reset_handle() issues a bus reset. ie: the bus reset is
taking place just like it happens for the first time during operational mode.
To denote the actual device state and the state we expect, a flag i2400m->alive
is introduced in addition to the existing flag i2400m->updown.
It's maintained with the same way for i2400m->updown but instead of reflecting
the actual state like i2400m->updown does, i2400m->alive maintains the state
we expect. i2400m->alive is set 1 just like whenever i2400m->updown is set 1.
Yet i2400m->alive remains 1 since we expect the device to be up all the time
until the driver is removed. See the doc for @alive in i2400m.h.
An enumeration I2400M_BUS_RESET_RETRIES is added to define the maximum number of
bus resets that a device reboot can retry.
A counter i2400m->bus_reset_retries is added to track how many bus resets
have been retried in one device reboot. If I2400M_BUS_RESET_RETRIES bus resets
were retried in this boot, we give up any further retrying so the device would enter
low power state. The counter i2400m->bus_reset_retries is incremented whenever
dev_reset_handle() is issuing a bus reset and is cleared to 0 when dev_start() is
successfully done, ie: a successful reboot.
Signed-off-by: Cindy H Kao <cindy.h.kao@intel.com>
This fix is to correct order of the handlers in the error path
of dev_start(). When i2400m_firmware_check fails, all the works done
before it should be released or cleared.
Signed-off-by: Cindy H Kao <cindy.h.kao@intel.com>
The race condition happens when the TX queue is accessed by
the TX work while the same TX queue is being destroyed because
a bus reset is triggered either by debugfs entry or simply
by failing waking up the device from WiMAX IDLE mode.
This fix is to prevent the TX queue from being accessed by
multiple threads
Signed-off-by: Cindy H Kao <cindy.h.kao@intel.com>
This patch increases the Tx buffer size so as to accommodate 12 payloads
of 1408 (1400 MTU 16 bytes aligned). Currently Tx buffer is 32 KiB which
is insufficient to accommodate 12 payloads of 1408 size.
This patch
- increases I2400M_TX_BUF_SIZE from 32KiB to 64KiB
- Adds a BUILD_BUG_ON if the calculated buffer size based
on the given MTU exceeds the I2400M_TX_BUF_SIZE.
Below is how we calculate the size of the Tx buffer.
Payload + 4 bytes prefix for each payload (1400 MTU 16 bytes boundary aligned)
= (1408 + sizeof(struct i2400m_pl_data_hdr)) * I2400M_TX_PLD_MAX
Adding 16 byte message header = + sizeof(struct i2400m_msg_hdr)
Aligning to 256 byte boundary
Total Tx buffer = (((((1408 + sizeof(struct i2400m_pl_data_hdr))
* I2400M_TX_PLD_MAX )+ sizeof(struct i2400m_msg_hdr))
/ 256) + 1) * 256 * 2
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
This patch moves I2400M_MAX_MTU enum defined in netdev.c to i2400m.h.
Follow up changes will make use of this value in other location,
thus requiring it to be moved to a global header file i2400m.h.
Signed-off-by: Prasanna S. Panchamukhi <prasannax.s.panchamukhi@intel.com>
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
i2400m_tx() routine was returning -ESHUTDOWN even when there was no Tx buffer
available. This patch fixes the i2400m_tx() to return -ESHUTDOWN only when
the device is down(i2400m->tx_buf is NULL) and also to return -ENOSPC
when there is no Tx buffer. Error seen in the kernel log.
kernel: i2400m_sdio mmc0:0001:1: can't send message 0x5606: -108
kernel: i2400m_sdio mmc0:0001:1: Failed to issue 'Enter power save'command: -108
Signed-off-by: Prasanna S.Panchamukhi <prasannax.s.panchamukhi@intel.com>
Get the device type from MODULE_BRINGUP_REQ command response.
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Move btmrvl hdev registration code out of btmrvl_add_card().
New function btmrvl_register_hdev() is added.
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
The MODULE_BRINGUP_REQ command response returns success with
either 0x00 or 0x0c.
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
This is mostly cleanup. There is only one caller and it just checks for
non-zero return values. Still "ret" should be int because we want to return
-EINVAL on errors.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Set the proper error(ENOMEM), instead of just return 0.
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Add prefetches of the skb and the next rx descriptor to speed up rx path.
Use prefetchw() for the skb [suggested by Eric Dumazet].
The rx descriptor is in skb->data which is mapped for streaming mode DMA.
Eric Dumazet pointed out that we should not prefetch the data before
dma_sync. So we prefetch only if dma_sync is no_op on the system.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
And turn on NETIF_F_GRO by default [requested by DaveM].
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set the napi member it to 0 explicitely.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the resource_size function instead of manually calculating the
resource size. This reduces the chance of introducing off-by-one
errors.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reset the PHY before first accessing it. Doing so, ensure that the PHY is
in a known good state before we read/write PHY registers. This fixes a
driver probe failure.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During Sx->S0 transitions, the interconnect between the MAC and PHY on
82577/82578 can remain in SMBus mode instead of transitioning to the
PCIe-like mode required during normal operation. Toggling the LANPHYPC
Value bit essentially resets the interconnect forcing it to the correct
mode.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on Andy's work, but I modified a lot.
Similar to the patch for bridge, this patch does:
1) implement the 2 methods to support netpoll for bonding;
2) modify netpoll during forwarding packets via bonding;
3) disable netpoll support of bonding when a netpoll-unabled device
is added to bonding;
4) enable netpoll support when all underlying devices support netpoll.
Cc: Andy Gospodarek <gospo@redhat.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: WANG Cong <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This whole patchset is for adding netpoll support to bridge and bonding
devices. I already tested it for bridge, bonding, bridge over bonding,
and bonding over bridge. It looks fine now.
To make bridge and bonding support netpoll, we need to adjust
some netpoll generic code. This patch does the following things:
1) introduce two new priv_flags for struct net_device:
IFF_IN_NETPOLL which identifies we are processing a netpoll;
IFF_DISABLE_NETPOLL is used to disable netpoll support for a device
at run-time;
2) introduce one new method for netdev_ops:
->ndo_netpoll_cleanup() is used to clean up netpoll when a device is
removed.
3) introduce netpoll_poll_dev() which takes a struct net_device * parameter;
export netpoll_send_skb() and netpoll_poll_dev() which will be used later;
4) hide a pointer to struct netpoll in struct netpoll_info, ditto.
5) introduce ->real_dev for struct netpoll.
6) introduce a new status NETDEV_BONDING_DESLAE, which is used to disable
netconsole before releasing a slave, to avoid deadlocks.
Cc: David Miller <davem@davemloft.net>
Cc: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: WANG Cong <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the PF acks a message from the VF the VF gets an interrupt. It
must cache the ack bit so that polling SW will not miss the ack. Also
avoid reading the message buffer on acks because that also will clear
the ack bit.
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver was calling the set Rx mode function for every multicast
filter set by the VF. When starting many VMs where each might have
multiple VLAN interfaces this would result in the function being
called hundreds or even thousands of times. This is unnecessary
for the case of the imperfect filters used in the MTA and has been
streamlined to be more efficient.
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver is unnecessarily writing values to VLAN control registers.
These writes already done elsewhere and are superfluous here.
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch reduces the number of skb cache misses in the
clean_tx_irq path, and results in an overall increase
in tx packet throughput.
Signed-off-by: Nicholas Nunley <nicholasx.d.nunley@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Accessing ks8851 companion eeprom permits modifying the ks8851 stored
MAC address.
Example how to change the MAC address using ethtool, to set the
01:23:45:67:89:AB MAC address:
$ echo "0:AB8976452301" | xxd -r > mac.bin
$ sudo ethtool -E eth0 magic 0x8870 offset 2 < mac.bin
Signed-off-by: Sebastien Jan <s-jan@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Low-level functions provide 16bits words read and write capability
to ks8851 companion eeprom.
Signed-off-by: Sebastien Jan <s-jan@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CCR register contains information on companion eeprom availability.
Signed-off-by: Sebastien Jan <s-jan@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Repeated calls to nv_rx_process in napi poll routine do not take
portion of budget that has been consumed in previous calls. Fix by
subtracting the number of packets processed.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Stop the queue when we add the packet that will fill it instead of dropping the packet
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We also reduce the high water mark to 1 so skbufs are not stranded for
long periods of time. Since we are cleaning after each packet, no
need to do it in the transmit path.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
