bgmac_open() calls phy_start() to initialize the PHY state machine,
which will set the interface's carrier state accordingly, no need to
force that as this could be conflicting with the PHY state determined by
PHYLIB.
Fixes: dd4544f054 ("bgmac: driver for GBit MAC core on BCMA bus")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver does not start the transmit queue in bgmac_open(). If the
queue was stopped prior to closing then re-opening the interface, we
would never be able to wake-up again.
Fixes: dd4544f054 ("bgmac: driver for GBit MAC core on BCMA bus")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We are checking for the Start of Frame bit in the ctl1 word, while this
bit is set in the ctl0 word instead. Read the ctl0 word and update the
check to verify that.
Fixes: 9cde94506e ("bgmac: implement scatter/gather support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since both CTAG and STAG rx acceleration must be enabled together, we
only need to check one feature flag (NETIF_F_HW_VLAN_CTAG_RX) before
calling __vlan_hwaccel_put_tag().
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The hardware can only be set to strip or not strip both the VLAN CTAG and
STAG. It cannot strip one and not strip the other. Add logic to
bnxt_fix_features() to toggle both feature flags when the user is toggling
one of them.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set the is_push flag in the software BD before the tx data is pushed to
the chip. It is possible to get the tx interrupt as soon as the tx data
is pushed. The tx handler will not handle the event properly if the
is_push flag is not set and it will crash.
Signed-off-by: Michael Chan <michael.chan@broadocm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since implementing VLAN filtering in commit 05cc5a39dd
("bnx2x: add vlan filtering offload") bnx2x refuses to add a VLAN while
the interface is down:
# ip link add link enp3s0f0 enp3s0f0_10 type vlan id 10
RTNETLINK answers: Bad address
and in dmesg (with bnx2x.debug=0x20):
bnx2x: [bnx2x_vlan_rx_add_vid:12941(enp3s0f0)]Ignoring VLAN
configuration the interface is down
Other drivers have no problem with this.
Fix this peculiar behavior in the following way:
- Accept requests to add/kill VID regardless of the device state.
Maintain the requested list of VIDs in the bp->vlan_reg list.
- If the device is up, try to configure the VID list into the hardware.
If we run out of VLAN credits or encounter a failure configuring an
entry, fall back to accepting all VLANs.
If we successfully configure all entries from the list, turn the
fallback off.
- Use the same code for reconfiguring VLANs during NIC load.
Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Acked-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
bnx2x_init_bp() allocates memory with bnx2x_alloc_mem_bp() so if we
fail later in bnx2x_init_one() we need to free this memory
with bnx2x_free_mem_bp() to avoid leakages. E.g. I'm observing memory
leaks reported by kmemleak when a failure (unrelated) happens in
bnx2x_vfpf_acquire().
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Acked-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch defines two new GSO definitions SKB_GSO_IPXIP4 and
SKB_GSO_IPXIP6 along with corresponding NETIF_F_GSO_IPXIP4 and
NETIF_F_GSO_IPXIP6. These are used to described IP in IP
tunnel and what the outer protocol is. The inner protocol
can be deduced from other GSO types (e.g. SKB_GSO_TCPV4 and
SKB_GSO_TCPV6). The GSO types of SKB_GSO_IPIP and SKB_GSO_SIT
are removed (these are both instances of SKB_GSO_IPXIP4).
SKB_GSO_IPXIP6 will be used when support for GSO with IP
encapsulation over IPv6 is added.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the weaker but more appropriate dma_rmb() to order the reading of
the completion ring.
Suggested-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current code is more complicated than necessary and can only report
unsupported SFP+ module if it is plugged in after the device is up.
Rename bnxt_port_module_event() to bnxt_get_port_module_status(). We
already have the current module_status in the link_info structure, so
just check that and report any unsupported SFP+ module status. Delete
the unnecessary last_port_module_event. Call this function at the
end of bnxt_open to report unsupported module already plugged in.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The len value in the hwrm error message is wrong. Use the properly adjusted
value in the variable len.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current code has 2 problems:
1. The maximum wait time is not long enough. It is about 60% of the
duration specified by the firmware. It is calling usleep_range(600, 800)
for every 1 msec we are supposed to wait.
2. The granularity of the delay is too coarse. Many simple firmware
commands finish in 25 usec or less.
We fix these 2 issues by multiplying the original 1 msec loop counter by
40 and calling usleep_range(25, 40) for each iteration.
There is also a second delay loop to wait for the last DMA word to
complete. This delay loop should be a very short 5 usec wait.
This change results in much faster bring-up/down time:
Before the patch:
time ip link set p4p1 up
real 0m0.120s
user 0m0.001s
sys 0m0.009s
After the patch:
time ip link set p4p1 up
real 0m0.030s
user 0m0.000s
sys 0m0.010s
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The chip supports 4K/8K/64K page sizes for the rings and we try to
match it to the CPU PAGE_SIZE. The current page size limits for the rings
are based on 4K/8K page size. If the page size is 64K, these limits are
too large. Reduce them appropriately.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add code to log a message during driver load indicating PCIe link
speed and width.
The log message will look like this:
bnxt_en 0000:86:00.0 eth0: PCIe: Speed 8.0GT/s Width x8
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to fetch the SFP EEPROM settings from the firmware
and display it via the ethtool -m command. We support SFP+ and QSFP
modules.
v2: Fixed a bug in bnxt_get_module_eeprom() found by Ben Hutchings.
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When there is only 1 MSI-X vector or in INTA mode, tx and rx pre-set
max channel parameters are shown incorrectly in ethtool -l. With only 1
vector, bnxt_get_max_rings() will return -ENOMEM. bnxt_get_channels
should check this return value, and set max_rx/max_tx to 0 if it is
non-zero.
Signed-off-by: Satish Baddipadige <sbaddipa@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The nf_conntrack_core.c fix in 'net' is not relevant in 'net-next'
because we no longer have a per-netns conntrack hash.
The ip_gre.c conflict as well as the iwlwifi ones were cases of
overlapping changes.
Conflicts:
drivers/net/wireless/intel/iwlwifi/mvm/tx.c
net/ipv4/ip_gre.c
net/netfilter/nf_conntrack_core.c
Signed-off-by: David S. Miller <davem@davemloft.net>
Add detection and recovery code when the hardware returned opaque value
does not match the expected consumer index. Once the issue is detected,
we skip the processing of all RX and LRO/GRO packets. These completion
entries are discarded without sending the SKB to the stack and without
producing new buffers. The function will be reset from a workqueue.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a rare hardware bug that can cause a bad opaque value in the RX
or TPA completion. When this happens, the hardware may have used the
same buffer twice for 2 rx packets. In addition, the driver will also
crash later using the bad opaque as the index into the ring.
The rx opaque value is predictable and is always monotonically increasing.
The workaround is to keep track of the expected next opaque value and
compare it with the one returned by hardware during RX and TPA start
completions. If they miscompare, we will not process any more RX and
TPA completions and exit NAPI. We will then schedule a workqueue to
reset the function.
This patch adds the logic to keep track of the next rx consumer index.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In netdevice.h we removed the structure in net-next that is being
changes in 'net'. In macsec.c and rtnetlink.c we have overlaps
between fixes in 'net' and the u64 attribute changes in 'net-next'.
The mlx5 conflicts have to do with vxlan support dependencies.
Signed-off-by: David S. Miller <davem@davemloft.net>
We recently had a system crash in the cnic module. Vmcore analysis confirmed
that "ip link up" was executed which failed due to an allocation failure
because of memory fragmentation. Futher analysis revealed that the cnic irq
vector was still allocated after the "ip link up" that failed. When
"ip link down" was executed it called free_msi_irqs() which crashed the system
because the cnic irq was still inuse.
PANIC: "kernel BUG at drivers/pci/msi.c:411!"
The code execution was:
cnic_netdev_event()
if (event == NETDEV_UP) {
.
.
▹ if (!cnic_start_hw(dev))
cnic_start_hw()
calls cnic_cm_open() which failed with -ENOMEM
cnic_start_hw() then took the err1 path:
err1:↩
cp->free_resc(dev);↩ <---- frees resources but not irq vector
pci_dev_put(dev->pcidev);↩
return err;↩
}↩
This returns control back to cnic_netdev_event() but now the cnic irq vector
is still allocated even although cnic_cm_open() failed. The next
"ip link down" while trigger the crash.
The cnic_start_hw() routine is not handling the allocation failure correctly.
Fix this by checking whether CNIC_DRV_STATE_HANDLES_IRQ flag is set indicating
that the hardware has been started in cnic_start_hw(). If it has then call
cp->stop_hw() which frees the cnic irq vector and cnic resources. Otherwise
just maintain the previous behaviour and free cnic resources.
I reproduced this by injecting an ENOMEM error into cnic_cm_alloc_mem()s return
code.
# ip link set dev enpX down
# ip link set dev enpX up <--- hit's allocation failure
# ip link set dev enpX down <--- crashes here
With this patch I confirmed there was no crash in the reproducer.
Signed-off-by: Jon Maxwell <jmaxwell37@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The multicast/all-multicast internal flags are not properly restored
after device reset. This could lead to unreliable multicast operations
after an ethtool configuration change for example.
Call bnxt_mc_list_updated() and setup the vnic->mask in bnxt_init_chip()
to fix the issue.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code determines if the next ring entry is valid before proceeding
further to read the rest of the entry. The CPU can re-order and read
the rest of the entry first, possibly reading a stale entry, if DMA
of a new entry happens right after reading it. This issue can be
readily seen on a ppc64 system, causing it to crash.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch assumes that the bnxt hardware will ignore existing IPv4/v6
header fields for length and checksum as well as the length and checksum
fields for outer UDP and GRE headers.
I have been told by Michael Chan that this is working. Though this might
be somewhat redundant for IPv6 as they are forcing the checksum to be
computed for all IPv6 frames that are offloaded. A follow-up patch may be
necessary in order to fix this as it is essentially mangling the outer IPv6
headers to add a checksum where none was requested.
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
net/ipv4/ip_gre.c
Minor conflicts between tunnel bug fixes in net and
ipv6 tunnel cleanups in net-next.
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fix spelling typos in printk from various part
of the codes.
Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
If PAGE_SIZE is bigger than BNXT_RX_PAGE_SIZE, that means the native CPU
page is bigger than the maximum length of the RX BD. Divide the page
into multiple 32K buffers for the aggregation ring.
Add an offset field in the bnxt_sw_rx_agg_bd struct to keep track of the
page offset of each buffer. Since each page can be referenced by multiple
buffer entries, call get_page() as needed to get the proper reference
count.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RX BD length field of this device is 16-bit, so the largest buffer
size is 65535. For LRO and GRO, we allocate native CPU pages for the
aggregation ring buffers. It won't work if the native CPU page size is
64K or bigger.
We fix this by defining BNXT_RX_PAGE_SIZE to be native CPU page size
up to 32K. Replace PAGE_SIZE with BNXT_RX_PAGE_SIZE in all appropriate
places related to the rx aggregation ring logic.
The next patch will add additional logic to divide the page into 32K
chunks for aggrgation ring buffers if PAGE_SIZE is bigger than
BNXT_RX_PAGE_SIZE.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only MSI-X can be used on a VF. The driver should fail initialization
if it cannot successfully enable MSI-X.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts were two cases of simple overlapping changes,
nothing serious.
In the UDP case, we need to add a hlist_add_tail_rcu()
to linux/rculist.h, because we've moved UDP socket handling
away from using nulls lists.
Signed-off-by: David S. Miller <davem@davemloft.net>
By using napi_complete_done(), we allow fine tuning of
/sys/class/net/ethX/gro_flush_timeout for higher GRO aggregation
efficiency for a Gbit NIC.
Check commit 24d2e4a507 ("tg3: use napi_complete_done()") for details.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Both bcm_sysport_tx_isr() and bcm_sysport_rx_isr() run in hard irq
context, we do not need to block irq again.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On 64bit kernels, device stats are 64bit wide, not 32bit.
Fixes: 1c1008c793 ("net: bcmgenet: add main driver file")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only core revisions older than 4 use BGMAC_CMDCFG_SR_REV0. This mainly
fixes support for BCM4708A0KF SoCs with Ethernet core rev 5 (it means
only some devices as most of BCM4708A0KF-s got core rev 4).
This was tested for regressions on BCM47094 which doesn't seem to care
which bit gets used.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes Ethernet on D-Link DIR-885L with BCM47094 SoC. Felix reported
similar fix was needed for his BCM4709 device (Buffalo WXR-1900DHP?).
I tested this for regressions on BCM4706, BCM4708A0 and BCM47081A0.
Cc: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add Byte Queue Limits (BQL) support to bcmgenet driver.
Signed-off-by: Petri Gynther <pgynther@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
bcmgenet_isr1() and bcmgenet_isr0() run in hard irq context,
we do not need to block irq again.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Petri Gynther <pgynther@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
By using napi_complete_done(), we allow fine tuning
of /sys/class/net/ethX/gro_flush_timeout for higher GRO aggregation
efficiency for a Gbit NIC.
Check commit 24d2e4a507 ("tg3: use napi_complete_done()") for details.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Petri Gynther <pgynther@google.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Petri Gynther <pgynther@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On some dual port cards, link speeds on both ports have to be compatible.
Firmware will inform the driver when a certain speed is no longer
supported if the other port has linked up at a certain speed. Add
logic to handle this event by logging a message and getting the
updated list of supported speeds.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some hypervisors (e.g. ESX) require the VF MAC address to be forwarded to
the PF for approval. In Linux PF, the call is not forwarded and the
firmware will simply check and approve the MAC address if the PF has not
previously administered a valid MAC address for this VF.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Let firmware know that the driver is giving up control of the link so that
it can be shutdown if no management firmware is running.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10GBaseT devices must autonegotiate to determine master/slave clocking.
Disallow forced speed in ethtool .set_settings() for these devices.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
dmadesc_set() is used for setting the Tx buffer DMA address, length,
and status bits on a Tx ring descriptor when a frame is being Tx'ed.
Always set the Tx buffer DMA address first, before updating the length
and status bits, i.e. giving the Tx descriptor to the hardware.
The reason this is a cleanup rather than a fix is that the hardware
won't transmit anything from a Tx ring until the TDMA producer index
has been incremented. As long as the dmadesc_set() writes complete
before the TDMA producer index write, life is good.
Signed-off-by: Petri Gynther <pgynther@google.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add frag_size = skb_frag_size(frag) and use it when needed.
Signed-off-by: Petri Gynther <pgynther@google.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. Readability: Move nr_frags assignment a few lines down in order
to bundle index -> ring -> txq calculations together.
2. Readability: Add parentheses around nr_frags + 1.
3. Minor fix: Stop the Tx queue and throw the error message only if
the Tx queue hasn't already been stopped.
Signed-off-by: Petri Gynther <pgynther@google.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If autoneg is off, we should always report the speed and duplex settings
even if it is link down so the user knows the current settings. The
unknown speed and duplex should only be used for autoneg when link is
down.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
