In a bidirectional forwarding test, we find that the best performance
is achieved by sending the TX completion interrupts from one NIC to a
CPU which shares an L2 cache with RX completion interrupts from the
other NIC. To facilitate this, add an option (through a module
parameter) to create separate channels for RX and TX completion with
separate IRQs when MSI-X is available.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is needed for recovery in case a PHY firmware upgrade is aborted.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The SFC4000 has strap pins indicating the presence of SPI flash and/or
EEPROM. These pins are also used for GPIO, and in some cases they may
be read wrongly at reset. However, on production boards it must boot
from one or the other device, so we can assume the boot device is
present and read the board config from there.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make falcon_spi_wait() ignore the write timer - it is only relevant to
write commands, it only works for the device that contains VPD, and it
might not be initialised properly at all.
Rename falcon_spi_fast_wait() to falcon_spi_wait_write(), reflecting
its use, and make it wait up to 10 ms (not 1 ms) since buffered writes
to EEPROM may take this long to complete.
Make both wait functions sleep instead of busy-waiting.
Replace wait for command completion at top of falcon_spi_cmd() with a
single poll; no command should be running when the function starts.
Correct some comments.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each reset is serialised by the rtnl_lock anyway, so there's no win
per-NIC.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A value of 0 means indefinite repetition (until interrupted).
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This was only ever needed for an FPGA version of Falcon.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set dummy monitor method for unrecognised boards.
Clean up board resources if efx_pci_probe_main() fails after board has
been initialised.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert driver to new net_device_ops. Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Randy Dunlap found that SFC_MTD was selected when sfc was built-in and
the MTD core was a module. Don't allow that combination.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order for the network device ops get_stats call to be immutable, the handling
of the default internal network device stats block has to be changed. Add a new
helper function which replaces the old use of internal_get_stats.
Note: change return code to make it clear that the caller should not
go changing the returned statistics.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to a hardware bug, the originally assigned range cannot reliably
be used for boot configuration and must not be modifiable through
ethtool.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The TX watchdog should trigger a reset, but a temperature/power alarm
should not as this is unlikely to solve the problem.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Add board monitoring to periodic work whenever link is down.
For SFE4001, report when a fault has caused the PHY to turn off.
For SFE4002, switch XFP PHY into low-power state in case of a fault.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The boot code that appears as a PCI expansion ROM on the SFC4000 is
stored in flash. Expose this as a standard MTD device to allow for
in-place upgrades.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Move flash and EEPROM partition boundary constants into spi.h and rename
them to be consistent.
Add a comment on the partitioning.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Due to a hardware bug, the originally assigned range cannot reliably
be used for boot configuration and must not be modifiable through
ethtool.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The generic packet receive code takes care of setting
netdev->last_rx when necessary, for the sake of the
bonding ARP monitor.
Drivers need not do it any more.
Some cases had to be skipped over because the drivers
were making use of the ->last_rx value themselves.
Signed-off-by: David S. Miller <davem@davemloft.net>
This converts pretty much everything to print_mac. There were
a few things that had conflicts which I have just dropped for
now, no harm done.
I've built an allyesconfig with this and looked at the files
that weren't built very carefully, but it's a huge patch.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
FALCON_SPI_MAX_LEN has type size_t while other SPI lengths have type
unsigned int. This results in warnings from min() on 64-bit
architectures where they are different. Add a cast to make it match.
From: Steve Hodgson <shodgson@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
For some buffers we use a starting offset of either NET_IP_ALIGN or 0
depending on whether we believe the architecture supports efficient
access to unaligned words. There is now a config macro specifying
whether this is the case, so check that rather than checking for
specific architectures.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This should avoid an interrupt storm, which has been observed in the
field with one faulty board.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This prevents speculative reading of the statistics before the
completion flag.
From: Neil Turton <nturton@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
mdio_clause45_links_ok() correctly checks efx_phy_mode_disabled(), so
tenxpress_link_ok() doesn't need to.
From: Steve Hodgson <shodgson@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Increase the potential retry count for RX flushes from 5 to 100.
Stop polling the RX_DESC_PTR_TBL to infer that a flush might have
happened. Instead absolutely rely on the flush events, unless bug 7803
applies (Falcon rev A only).
To keep things quick, request flushes for every TX and RX queue up
front, and match up the events to requests.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
On some boards 10Xpress feeds a 156 MHz clock to the Falcon XMAC. MAC
statistics DMA can fail while this clock is stopped during a PHY reset.
From: Steve Hodgson <shodgson@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
There was a bug in XAUI synchronisation in early 10Xpress firmware
versions. This is fixed in released firmware and we do not need to
work around it.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Match pci_request_region() with pci_release_region(), not
release_mem_region().
From: Steve Hodgson <shodgson@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
When !port_enabled, defer the write to reconfigure_mac_wrapper.
Whilst here, simplify the logic now that efx_start_port() always calls
efx_reconfigure_port().
From: Steve Hodgson <shodgson@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This was originally a kludge to fix broken locking, which has since
been fixed properly.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Include PMA/PMD in loopback self-tests as intended.
Add NVRAM checksum validation and include it in self-tests.
Add register self-tests.
Run PHY self-tests where available.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This will support register self-tests.
From: Steve Hodgson <shodgson@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Move more code from efx_reset() into efx_reset_down() and efx_reset_up().
Stop propagating MAC/PHY setting failures from efx_reset_down() and
efx_reset_up() as these should not be fatal.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Some functions return an error code which is always 0. Change their
return types to void and simplify their callers accordingly.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Update comment on the dummy operation implementations.
Line up the board operation initialisers.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This allows updating PHY firmware for one interface without removing
all other interfaces handled by the driver.
Replace tx_disabled flags and 10Xpress status enumeration with flags in
enum efx_phy_mode.
Prevent an interface from being brought up while in PHY flash mode.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Make efx_process_channel() and falcon_process_eventq() return the
number of packets received rather than updating the quota, consistent
with new NAPI.
Since channels and RX queues are mapped one-to-one, remove return
value from falcon_handle_rx_event() and add a warning for events
with the wrong RX queue number.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Currently each channel can have at most one RX queue, so go straight to
that one.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
We have long since given up doing RSS on Falcon A1 and therefore we
would always write the default value of 0.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Rename efx_nic::rss_queues to the more obvious n_rx_queues
Remove efx_rx_queue::used and other stuff that's redundant with it.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
efx_channel::has_interrupt is redundant with efx_channel::used_flags.
Remove efx_test_eventq() because it is now obviously unreachable.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Extend the SPI device setup code to support this.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
gcc will automatically inline static functions with only one caller, and
may inline other functions depending on the kernel configuration and size
of the intermediate code.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Squashed nested structures.
Renamed remaining_len to out_len, ifc.len to in_len, header_length to
header_len.
Moved ipv4_id into the group of output variables where it belongs.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Replace (cond ? 1 : 0) with cond or !!cond as appropriate, and
(cond ? 0 : 1) with !cond.
Remove some redundant boolean temporaries.
Rename one field that looks like a flag but isn't.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Remove unmap_addr since it can be calculated from dma_addr, len and
unmap_len. This saves 4-16 bytes.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Add efx_poll_loopback() function to test for successful completion of test.
Change efx_test_loopback() to end the test after 1 ms if
efx_poll_loopback() indicates success, and otherwise to wait for 100 ms
as before.
While we're here, rename efx_{rx,tx}_loopback() to
efx_{begin,end}_loopback() which more accurately reflect their
purpose.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Removed log messages that are redundant with calling functions.
Fixed bitwise or-ing of return codes.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This was making noise during 10Xpress self-test.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
These were clearly bogus.
From: Steve Hodgson <shodgson@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Exclude assumed size of RX control frames from rx_bad_bytes.
Exclude assumed size of TX control frames from tx_good_bytes for
consistency with rx_good_bytes.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
We believed that some valid SNAP frames were being marked as invalid.
In fact this is not the case and no workaround is needed.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Currently the board initialisation includes 2 delays of 1 second each.
Usually it is unnecessary to wait that long, so check before doing so.
Correct some of the comments and log messages while we're here.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
efx_test_loopback() used "|" to combine the results of the RX and TX
phases. If both phases fail with different error codes, this results
in a bogus error code.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Checksum generation is an attribute of our hardware TX queues, not TX
descriptors. We previously used a single queue and turned checksum
generation on or off as requested through ethtool. However, this can
result in regenerating checksums in raw packets that should not be
modified. We now create 2 hardware TX queues with checksum generation
on or off. They are presented to the net core as one queue since it
does not know how to select between them.
The self-test verifies that a bad checksum is unaltered on the queue
with checksum generation off.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This means the compiler doesn't need to use real division instructions.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Use of the net_device::priv field is deprecated.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Add per-device dma_mapping_ops support for CONFIG_X86_64 as POWER
architecture does:
This enables us to cleanly fix the Calgary IOMMU issue that some devices
are not behind the IOMMU (http://lkml.org/lkml/2008/5/8/423).
I think that per-device dma_mapping_ops support would be also helpful for
KVM people to support PCI passthrough but Andi thinks that this makes it
difficult to support the PCI passthrough (see the above thread). So I
CC'ed this to KVM camp. Comments are appreciated.
A pointer to dma_mapping_ops to struct dev_archdata is added. If the
pointer is non NULL, DMA operations in asm/dma-mapping.h use it. If it's
NULL, the system-wide dma_ops pointer is used as before.
If it's useful for KVM people, I plan to implement a mechanism to register
a hook called when a new pci (or dma capable) device is created (it works
with hot plugging). It enables IOMMUs to set up an appropriate
dma_mapping_ops per device.
The major obstacle is that dma_mapping_error doesn't take a pointer to the
device unlike other DMA operations. So x86 can't have dma_mapping_ops per
device. Note all the POWER IOMMUs use the same dma_mapping_error function
so this is not a problem for POWER but x86 IOMMUs use different
dma_mapping_error functions.
The first patch adds the device argument to dma_mapping_error. The patch
is trivial but large since it touches lots of drivers and dma-mapping.h in
all the architecture.
This patch:
dma_mapping_error() doesn't take a pointer to the device unlike other DMA
operations. So we can't have dma_mapping_ops per device.
Note that POWER already has dma_mapping_ops per device but all the POWER
IOMMUs use the same dma_mapping_error function. x86 IOMMUs use device
argument.
[akpm@linux-foundation.org: fix sge]
[akpm@linux-foundation.org: fix svc_rdma]
[akpm@linux-foundation.org: build fix]
[akpm@linux-foundation.org: fix bnx2x]
[akpm@linux-foundation.org: fix s2io]
[akpm@linux-foundation.org: fix pasemi_mac]
[akpm@linux-foundation.org: fix sdhci]
[akpm@linux-foundation.org: build fix]
[akpm@linux-foundation.org: fix sparc]
[akpm@linux-foundation.org: fix ibmvscsi]
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Avi Kivity <avi@qumranet.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Using multiple cores in the same package to handle received traffic
does not appear to provide a performance benefit. Therefore use CPU
topology information to count CPU packages and use that as the default
number of RX queues and interrupts. We rely on interrupt balancing to
spread the interrupts across packages.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This avoids deadlock in case a reset is triggered during self-test.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
As recommended by Jean Delvare:
- Increase timeout to 50 ms
- Leave adapter class clear so that unwanted drivers do not probe our bus
- Use strlcpy() for name initialisation
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Now that we have a specific lock to protect the network
device unicast and multicast lists, remove extraneous
grabs of the TX lock in cases where the code only needs
address list protection.
Signed-off-by: David S. Miller <davem@davemloft.net>
Add netif_addr_{lock,unlock}{,_bh}() helpers.
Use them to protect operations that operate on or read
the network device unicast and multicast address lists.
Also use them in cases where the code simply wants to
block calls into the driver's ->set_rx_mode() and
->set_multicast_list() methods.
Signed-off-by: David S. Miller <davem@davemloft.net>
RX queue flush can fail if traffic continues to arrive. Recover by
performing an invisible reset.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Remove our own implementation of I2C bit-banging.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This is normal when the external link is down so don't report it as an error.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This driver has diverged from the out-of-tree driver to which the version
number originally applied. It should be identified primarily by kernel
version.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Also change type of efx_nic_type::max_dma_mask to u64, matching
pci_dma_supported() parameter type.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
These comments have been revised in response to questions raised by Andrew
Morton in <20080501120858.207b6dd6.akpm@linux-foundation.org>.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This should make the driver work on 32-bit systems with 64-bit PCI support.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Merged it into its only caller, efx_init_channels().
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Add a set of self-tests accessible thorugh ethtool.
Add hardware loopback and TX disable control code to support them.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
An RX_RESET event can be reported either as a global or as a driver event.
We were counting only global events.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Fall-through is expected outside a switch statement.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The 10Xpress PHY supports flash upgrades through MDIO, but needs to be
put in upgrade mode at power-up. This adds a module parameter and other
logic to support that.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The SFC4000 controller does not have hardware support for TSO, and the
core GSO code incurs a high cost in allocating and freeing skbs. This
TSO implementation uses lightweight packet header structures and is
substantially faster.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The driver supports the 10Xpress PHY and XFP modules on our reference
designs SFE4001 and SFE4002 and the SMC models SMC10GPCIe-XFP and
SMC10GPCIe-10BT.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Neil Turton