Suresh Reddy [Thu, 25 May 2017 02:24:38 +0000 (22:24 -0400)]
be2net: Fix UE detection logic for BE3
On certain platforms BE3 chips may indicate spurious UEs (unrecoverable
error). Because of the UE detection logic was disabled in the driver
for BE3 chips. Because of this, even in cases of a real UE,
a failover will not occur. This patch re-enables UE detection on BE3
and if a UE is detected, reads the POST register. If the POST register,
reports either a FAT_LOG_STATE or a ARMFW_UE, then it means that a valid
UE occurred in the chip.
Signed-off-by: Suresh Reddy <suresh.reddy@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Kirtikar Kashyap <kirtikar.kashyap@oracle.com> Reviewed-by: Reviewed-by: Jack Vogel <jack.vogel@oracle.com>
Vlad Yasevich [Tue, 23 May 2017 17:38:42 +0000 (13:38 -0400)]
be2net: Fix offload features for Q-in-Q packets
At least some of the be2net cards do not seem to be capabled
of performing checksum offload computions on Q-in-Q packets.
In these case, the recevied checksum on the remote is invalid
and TCP syn packets are dropped.
This patch adds a call to check disbled acceleration features
on Q-in-Q tagged traffic.
Signed-off-by: Karim Eshapa <karim.eshapa@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Kirtikar Kashyap <kirtikar.kashyap@oracle.com> Reviewed-by: Reviewed-by: Jack Vogel <jack.vogel@oracle.com>
Signed-off-by: Suresh Reddy <suresh.reddy@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Kirtikar Kashyap <kirtikar.kashyap@oracle.com> Reviewed-by: Reviewed-by: Jack Vogel <jack.vogel@oracle.com>
Ivan Vecera [Tue, 31 Jan 2017 19:01:31 +0000 (20:01 +0100)]
be2net: fix initial MAC setting
Recent commit 34393529163a ("be2net: fix MAC addr setting on privileged
BE3 VFs") allows privileged BE3 VFs to set its MAC address during
initialization. Although the initial MAC for such VFs is already
programmed by parent PF the subsequent setting performed by VF is OK,
but in certain cases (after fresh boot) this command in VF can fail.
The MAC should be initialized only when:
1) no MAC is programmed (always except BE3 VFs during first init)
2) programmed MAC is different from requested (e.g. MAC is set when
interface is down). In this case the initial MAC programmed by PF
needs to be deleted.
The adapter->dev_mac contains MAC address currently programmed in HW so
it should be zeroed when the MAC is deleted from HW and should not be
filled when MAC is set when interface is down in be_mac_addr_set() as
no programming is performed in this case.
Example of failure without the fix (immediately after fresh boot):
be2net 0000:01:00.0 eth0: Link is Up
...
be2net 0000:01:04.0: Emulex OneConnect(be3): VF port 0
Eric Dumazet [Mon, 30 Jan 2017 16:22:01 +0000 (08:22 -0800)]
drivers: net: generalize napi_complete_done()
napi_complete_done() allows to opt-in for gro_flush_timeout,
added back in linux-3.19, commit 3b47d30396ba
("net: gro: add a per device gro flush timer")
This allows for more efficient GRO aggregation without
sacrifying latencies.
Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Kirtikar Kashyap <kirtikar.kashyap@oracle.com> Reviewed-by: Reviewed-by: Jack Vogel <jack.vogel@oracle.com>
Ivan Vecera [Fri, 13 Jan 2017 21:38:29 +0000 (22:38 +0100)]
be2net: fix MAC addr setting on privileged BE3 VFs
During interface opening MAC address stored in netdev->dev_addr is
programmed in the HW with exception of BE3 VFs where the initial
MAC is programmed by parent PF. This is OK when MAC address is not
changed when an interfaces is down. In this case the requested MAC is
stored to netdev->dev_addr and later is stored into HW during opening.
But this is not done for all BE3 VFs so the NIC HW does not know
anything about this change and all traffic is filtered.
This is the case of bonding if fail_over_mac == 0 where the MACs of
the slaves are changed while they are down.
The be2net behavior is too restrictive because if a BE3 VF has
the FILTMGMT privilege then it is able to modify its MAC without
any restriction.
To solve the described problem the driver should take care about these
privileged BE3 VFs so the MAC is programmed during opening. And by
contrast unpriviled BE3 VFs should not be allowed to change its MAC
in any case.
Ivan Vecera [Fri, 6 Jan 2017 20:59:30 +0000 (21:59 +0100)]
be2net: fix unicast list filling
The adapter->pmac_id[0] item is used for primary MAC address but
this is not true for adapter->uc_list[0] as is assumed in
be_set_uc_list(). There are N UC addresses copied first from net_device
to adapter->uc_list[1..N] and then N UC addresses from
adapter->uc_list[0..N-1] are sent to HW. So the last UC address is never
stored into HW and address 00:00:00:00;00:00 (from uc_list[0]) is used
instead.
Ivan Vecera [Fri, 6 Jan 2017 19:30:02 +0000 (20:30 +0100)]
be2net: fix accesses to unicast list
Commit 988d44b "be2net: Avoid redundant addition of mac address in HW"
introduced be_dev_mac_add & be_uc_mac_add helpers that incorrectly
access adapter->uc_list as an array of bytes instead of an array of
be_eth_addr. Consequently NIC is not filled with valid data so unicast
filtering is broken.
drivers/net/ethernet/emulex/benet/be_main.c:47:25: warning:
symbol 'be_err_recovery_workq' was not declared. Should it be static?
drivers/net/ethernet/emulex/benet/be_main.c:63:25: warning:
symbol 'be_wq' was not declared. Should it be static?
be2net: Avoid redundant addition of mac address in HW
If a mac address is added to the uc_list and later the same mac address
is added via ndo_set_mac_address() or vice versa, the driver does not
detect this condition and tries to add it again. This results in a mac
address collision error when the FW rejects it.
Fix this by checking if the given mac address is present in uc_list while
setting the device mac address and vice versa. Similarly skip deletion if
the address is still in use in the other form.
be2net: Support UE recovery in BEx/Skyhawk adapters
This patch supports recovery from UEs caused due to Transient Parity
Errors (TPE), in BE2, BE3 and Skyhawk adapters. This change avoids
system reboot when such errors occur. The driver recovers from these
errors such that the adapter resumes full operational status as prior
to the UE.
Following is the list of changes in the driver to support this:
o The driver registers its UE recoverable capability with ARM FW at init
time. This also allows the driver to know if the feature is supported in
the FW.
o As the UE recovery requires precise time bound processing, the driver
creates its own error recovery work queue with a single worker thread (per
module, shared across functions).
o Each function runs an error detection task at an interval of 1 second as
required by the FW. The error detection logic already exists for BEx/SH,
but it now runs in the context of a separate worker thread.
o When an error is detected by the task, if it is recoverable, the PF0
driver instance initiates a soft reset, while other PF driver instances
wait for the reset to complete and the chip to become ready. Once
the chip is ready, all driver instances including PF0, resume to
reinitialize the respective functions.
o The PF0 driver checks for some recovery criteria, to determine if the
recovery can be initiated. If the criteria is not met, the PF0 driver does
not initiate a soft reset, it retains the existing behavior to stop
further processing and requires a reboot to get the chip to operational
state again.
o To allow each function to share the workq, while also making progress in
its recovery process, a per-function recovery state machine is used.
The per-function tasks avoid blocking operations like msleep() while in
this state machine (until reinit state) and instead reschedule for the
required delay.
o With these changes, the existing error recovery code for Lancer also
runs in the context of the new worker thread.
be2net: replace polling with sleeping in the FW completion path
The ndo_set_rx_mode() and ndo_add/del_vxlan_port() calls may be called with
BHs disabled. The driver currently issues the required cmds to the FW in
these contexts and polls on completions from the FW, while BHs remain
disabled. This can cause either packet loss or packet reception to be
delayed on that CPU.
This patch defers processing of the above cmds to a separate workqueue.
With this change, FW cmds are now issued only in process context.
Now that the FW cmds are issued only in process context, they can sleep
waiting for a completion instead of polling. All the spin_lock_bh(mcc_lock)
calls are now replaced with mutex calls.
Also a new rx_filter_lock is now needed to protect the RX filtering fields
like vids[] between be_vlan_add/rem_vid() and __be_set_rx_mode() contexts.
Sathya Perla [Wed, 22 Jun 2016 12:54:54 +0000 (08:54 -0400)]
be2net: support asymmetric rx/tx queue counts
be2net so far supported creation of RX/TX queues only in pairs.
On configs where rx and tx queue counts are different, creation of only
the lesser number of queues has been supported.
This patch now allows a combination of RX/TX-only channels along with
combined channels. N TX-queues and M RX-queues can be created with the
following cmds:
ethtool -L ethX combined N rx M-N (when N < M)
ethtool -L ethX combined M tx N-M (when M < N)
Setting both RX-only and TX-only channels is still not supported.
It is mandatory to create atleast one combined channel.
Eric Dumazet [Wed, 15 Mar 2017 20:21:28 +0000 (13:21 -0700)]
net: properly release sk_frag.page
I mistakenly added the code to release sk->sk_frag in
sk_common_release() instead of sk_destruct()
TCP sockets using sk->sk_allocation == GFP_ATOMIC do no call
sk_common_release() at close time, thus leaking one (order-3) page.
iSCSI is using such sockets.
Fixes: 5640f7685831 ("net: use a per task frag allocator") Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 26409538
To handle netpoll properly, the driver must only handle TX packets
during NAPI. Handling RX events cause warnings and errors in
netpoll mode. The ndo_poll_controller() method should call
napi_schedule() directly so that a NAPI weight of zero will be used
during netpoll mode.
The bnxt_en driver supports 2 ring modes: combined, and separate rx/tx.
In separate rx/tx mode, the ndo_poll_controller() method will only
process the tx rings. In combined mode, the rx and tx completion
entries are mixed in the completion ring and we need to drop the rx
entries and recycle the rx buffers.
Add a function bnxt_force_rx_discard() to handle this in netpoll mode
when we see rx entries in combined ring mode.
Reported-by: Calvin Owens <calvinowens@fb.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 2270bc5da34979454e6f2eb133d800b635156174) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
When we get a TPA_END completion to handle a completed LRO packet, it
is possible that hardware would indicate errors. The current code is
not checking for the error condition. Define the proper error bits and
the macro to check for this error and abort properly.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 69c149e2e39e8d66437c9034bb4926ef2c1f7c23) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
We need to write the doorbell if BQL has stopped the queue and
skb->xmit_more is set. Otherwise it is possible for the tx queue to
rot and cause tx timeout.
Fixes: 4d172f21cefe ("bnxt_en: Implement xmit_more.") Suggested-by: Yuval Mintz <yuval.mintz@cavium.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit ffe406457753a7ca2061ecc8c4d3971623066911) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
In the existing code, the local variable sh is hardcoded to true to
calculate default rings for shared ring configuration. It is better
to have the caller determine the value of sh.
Reported-by: Gustavo A. R. Silva <garsilva@embeddedor.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 702c221ca64060b81af4461553be19cba275da8b) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Do not write the TX doorbell if skb->xmit_more is set unless the TX
queue is full.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 4d172f21cefe896df8477940269b8d52129f8c87) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Older chips require the doorbells to be written twice, but newer chips
do not. Add a new common function bnxt_db_write() to write all
doorbells appropriately depending on the chip. Eliminating the extra
doorbell on newer chips has a significant performance improvement
on pktgen.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 434c975a8fe2f70b70ac09ea5ddd008e0528adfa) Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
drivers/net/ethernet/broadcom/bnxt/bnxt.c
drivers/net/ethernet/broadcom/bnxt/bnxt.h
drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Add additional chip definitions and macros for all supported chips.
Add a new macro BNXT_CHIP_P4_PLUS for the newer generation of chips and
use the macro to properly determine the features supported by these
newer chips.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 3284f9e1ab505b41fa604c81e4b3271c6b88cdcb) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
When bnxt_en gets a PCI shutdown call, we need to have a new callback
to inform the RDMA driver to do proper shutdown and removal.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 0efd2fc65c922dff207ff10a776a7a33e0e3c7c5) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Signed-off-by: Deepak Khungar <deepak.khungar@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit c7ef35eb0c8d0b58d2d5ae5be599e6aa730361b2) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
The new short message format is used on the new BCM57454 VFs. Each
firmware message is a fixed 16-byte message sent using the standard
firmware communication channel. The short message has a DMA address
pointing to the legacy long firmware message.
Signed-off-by: Deepak Khungar <deepak.khungar@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit e605db801bdeb9d94cccbd4a2f641030067ef008) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Otherwise, all the host based DCBX settings from lldpad will fail if the
firmware DCBX agent is running.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit f667724b99ad1afc91f16064d8fb293d2805bd57) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
In the current code, bnxt_dcb_init() is called too early before we
determine if the firmware DCBX agent is running or not. As a result,
we are not setting the DCB_CAP_DCBX_HOST and DCB_CAP_DCBX_LLD_MANAGED
flags properly to report to DCBNL.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 87fe603274aa9889c05cca3c3e45675e1997cb13) Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
drivers/net/ethernet/broadcom/bnxt/bnxt.c Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Shannon Nelson [Tue, 4 Apr 2017 18:41:35 +0000 (11:41 -0700)]
bnxt: add dma mapping attributes
On the SPARC platform we need to use the DMA_ATTR_WEAK_ORDERING
attribute in our dma mapping in order to get the expected performance
out of the receive path. Setting this boosts a simple iperf receive
session from 2 Gbe to 23.4 Gbe.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com> Reviewed-by: Tushar Dave <tushar.n.dave@oracle.com> Reviewed-by: Tom Saeger <tom.saeger@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com> Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
drivers/net/ethernet/broadcom/bnxt/bnxt.c Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
We have the number of longs, but we need to calculate the number of
bytes required.
Fixes: c0c050c58d84 ("bnxt_en: New Broadcom ethernet driver.") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Acked-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit ac45bd93a5035c2f39c9862b8b6ed692db0fdc87) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
This change restricts the PF in multi-host mode from setting any port
level PHY configuration. The settings are controlled by firmware in
Multi-Host mode.
Signed-off-by: Deepak Khungar <deepak.khungar@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 9e54e322ded40f424dcb5a13508e2556919ce12a) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Check the additional flag in bnxt_hwrm_func_qcfg() before allowing
DCBX to be done in host mode.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 7d63818a35851cf00867248d5ab50a8fe8df5943) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Added support for 100G link speed reporting for Broadcom BCM57454
ASIC in ethtool command.
Signed-off-by: Deepak Khungar <deepak.khungar@broadcom.com> Signed-off-by: Ray Jui <ray.jui@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 38a21b34aacd4db7b7b74c61afae42ea6718448d) Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
The .ndo_get_vf_config() is returning the wrong qos attribute. Fix
the code that checks and reports the qos and spoofchk attributes. The
BNXT_VF_QOS and BNXT_VF_LINK_UP flags should not be set by default
during init. time.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit f0249056eaf2b9a17b2b76a6e099e9b7877e187d) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
When the driver gets the RoCE app priority set/delete call through DCBNL,
the driver will send the information to the firmware to set up the
priority VLAN tag for RDMA traffic.
[ New version using the common ETH_P_IBOE constant in if_ether.h ]
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit a82fba8dbfb522bd19b1644bf599135680fd0122) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
The current code enables up to the maximum MSIX vectors in the PCIE
config space without considering the max completion rings available.
An MSIX vector is only useful when it has an associated completion
ring, so it is better to cap it.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 68a946bb81e07ed0e59a99e0c068d091ed42cc1b) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 67fea463fd873492ab641459a6d1af0e9ea3c9ce) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
It is necessary to disable autoneg before enabling PHY loopback,
otherwise link won't come up.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 91725d89b97acea168a94c577d999801c3b3bcfb) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
The mac loopback self test operates in polling mode. To support that,
we need to add functions to open and close the NIC half way. The half
open mode allows the rings to operate without IRQ and NAPI. We
use the XDP transmit function to send the loopback packet.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit f7dc1ea6c4c1f31371b7098d6fae0d49dc6cdff1) Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Add the basic infrastructure and only firmware tests initially.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit eb51365846bc418687af4c4f41b68b6e84cdd449) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Add suspend/resume callbacks using the newer dev_pm_ops method.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit f65a2044a8c988adf16788c51c04ac10dbbdb494) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
And add functions to set and free magic packet filter.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 5282db6c794fed3ea8b399bc5305c4078e084f7b) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 8e202366dd752564d7f090ba280cc51cbf7bbbd9) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Add pci shutdown method to put device in the proper WoL and power state.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit d196ece740bf337aa25731cd8cb44660a2a227dd) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Add code to driver probe function to check if the device is WoL capable
and if Magic packet WoL filter is currently set.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit c1ef146a5bd3b286d5c3eb2c9f631b38647c76d3) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Signed-off-by: Deepak Khungar <deepak.khungar@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 8eb992e876a88de7539b1b9e132dd171d865cd2f) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
In bnxt_free_rx_skbs(), which is called to free up all RX buffers during
shutdown, we need to unmap the page if we are running in XDP mode.
Fixes: c61fb99cae51 ("bnxt_en: Add RX page mode support.") Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 3ed3a83e3f3871c57b18cef09b148e96921236ed) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Conflicts:
drivers/net/ethernet/broadcom/bnxt/bnxt.c
Signed-off-by: Sankar Patchineelam <sankar.patchineelam@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 23e12c893489ed12ecfccbf866fc62af1bead4b0) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Net device reset can fail when the h/w or f/w is in a bad state.
Subsequent netdevice open fails in bnxt_hwrm_stat_ctx_alloc().
The cleanup invokes bnxt_hwrm_resource_free() which inturn
calls bnxt_disable_int(). In this routine, the code segment
if (ring->fw_ring_id != INVALID_HW_RING_ID)
BNXT_CP_DB(cpr->cp_doorbell, cpr->cp_raw_cons);
results in NULL pointer dereference as cpr->cp_doorbell is not yet
initialized, and fw_ring_id is zero.
The fix is to initialize cpr fw_ring_id to INVALID_HW_RING_ID before
bnxt_init_chip() is invoked.
Signed-off-by: Sankar Patchineelam <sankar.patchineelam@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 2247925f0942dc4e7c09b1cde45ca18461d94c5f) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
In some situations, the firmware will return 0 for autoneg supported
speed. This may happen if the firmware detects no SFP module, for
example. The driver should ignore this so that we don't end up with
an invalid autoneg setting with nothing advertised. When SFP module
is inserted, we'll get the updated settings from firmware at that time.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 520ad89a54edea84496695d528f73ddcf4a52ea4) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Set DCB_CAP_DCBX_HOST capability flag only if the firmware LLDP agent
is not running.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit bc39f885a9c3bdbff0a96ecaf07b162a78eff6e4) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
If we call bnxt_reset_task() due to tx timeout, we should call
bnxt_ulp_stop() to inform the RDMA driver about the error and the
impending reset.
Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit b386cd362ffea09d05c56bfa85d104562e860647) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
The firmware call to do function reset is done too late. It is causing
the rings that have been reserved to be freed. In NPAR mode, this bug
is causing us to run out of rings.
Fixes: 391be5c27364 ("bnxt_en: Implement new scheme to reserve tx rings.") Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 3c2217a675bac22afb149166e0de71809189850d) Signed-off-by: Brian Maly <brian.maly@oracle.com> Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Dave Carroll [Thu, 22 Jun 2017 18:44:09 +0000 (12:44 -0600)]
Initialize fiblink list head during fib initialization
The fiblink pointer is used extensively for the sync mode driver, and has never been initialized
Initialize this during the fib_setup operation, and for copied fibs
Dave Carroll [Fri, 16 Jun 2017 20:59:27 +0000 (14:59 -0600)]
aacraid: Update scsi_host_template to use tagged commands
Most of the non-multi-queue scsi drivers were updated to include
.use_blk_tags in the scsi_host_template, however the aacraid driver
was left out. This will cause the inbox driver to fail to
allocate a tagged fib.
Update the scsi_host_template to include .use_blk_tags
Signed-off-by: Dave Carroll <david.carroll@microsemi.com> Signed-off-by: Kirtikar Kashyap <kirtikar.kashyap@oracle.com> Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
syzkaller found a way to trigger double frees from ip_mc_drop_socket()
It turns out that leave a copy of parent mc_list at accept() time,
which is very bad.
Very similar to commit 8b485ce69876 ("tcp: do not inherit
fastopen_req from parent")
Initial report from Pray3r, completed by Andrey one.
Thanks a lot to them !
Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Pray3r <pray3r.z@gmail.com> Reported-by: Andrey Konovalov <andreyknvl@google.com> Tested-by: Andrey Konovalov <andreyknvl@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
net/ipv4/inet_connection_sock.c
When ring buf full, hw queue will be stopped. While blkif interrupt consume
request and make free space in ring buf, hw queue will be started again.
But since start queue is protected by spin lock while stop not, that will
cause a race.
If ring buf is made empty in this case, interrupt will never come, then the
hw queue will be stopped forever, all processes waiting for the pending io
in the queue will hung.
Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com> Reviewed-by: Ankur Arora <ankur.a.arora@oracle.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Acked-by: Roger Pau Monné <roger.pau@citrix.com>
blk_mq_update_nr_hw_queues() used to remap hardware queues, which is the
behavior that drivers expect. However, commit 4e68a011428a changed
blk_mq_queue_reinit() to not remap queues for the case of CPU
hotplugging, inadvertently making blk_mq_update_nr_hw_queues() not remap
queues as well. This breaks, for example, NBD's multi-connection mode,
leaving the added hardware queues unused. Fix it by making
blk_mq_update_nr_hw_queues() explicitly remap the queues.
Fixes: 4e68a011428a ("blk-mq: don't redistribute hardware queues on a CPU hotplug event") Reviewed-by: Keith Busch <keith.busch@intel.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Sagi Grimberg <sagi@grimberg.me> Signed-off-by: Omar Sandoval <osandov@fb.com> Signed-off-by: Jens Axboe <axboe@fb.com>
(cherry picked from commit ebe8bddb6e30d7a02775b9972099271fc5910f37)
Conflicts:
block/blk-mq.c
Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com> Reviewed-by: Jack Vogel <jack.vogel@oracle.com>
Currently blk-mq will totally remap hardware context when a CPU hotplug
even happened, which causes major havoc for drivers, as they are never
told about this remapping. E.g. any carefully sorted out CPU affinity
will just be completely messed up.
The rebuild also doesn't really help for the common case of cpu
hotplug, which is soft onlining / offlining of cpus - in this case we
should just leave the queue and irq mapping as is. If it actually
worked it would have helped in the case of physical cpu hotplug,
although for that we'd need a way to actually notify the driver.
Note that drivers may already be able to accommodate such a topology
change on their own, e.g. using the reset_controller sysfs file in NVMe
will cause the driver to get things right for this case.
With the rebuild removed we will simplify retain the queue mapping for
a soft offlined CPU that will work when it comes back online, and will
map any newly onlined CPU to queue 0 until the driver initiates
a rebuild of the queue map.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <keith.busch@intel.com> Signed-off-by: Jens Axboe <axboe@fb.com>
(cherry picked from commit 4e68a011428af3211facd932b4003b3fa3ef4faa)
Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com> Reviewed-by: Jack Vogel <jack.vogel@oracle.com>
Signed-off-by: Manish Rangankar <manish.rangankar@cavium.com> Signed-off-by: Chad Dupuis <chad.dupuis@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Brian Maly <brian.maly@oracle.com>
Without CONFIG_DEBUG_FS, we run into a link error:
drivers/scsi/qedi/qedi_iscsi.o: In function `qedi_ep_poll':
qedi_iscsi.c:(.text.qedi_ep_poll+0x134): undefined reference to `do_not_recover'
drivers/scsi/qedi/qedi_iscsi.o: In function `qedi_ep_disconnect':
qedi_iscsi.c:(.text.qedi_ep_disconnect+0x36c): undefined reference to `do_not_recover'
drivers/scsi/qedi/qedi_iscsi.o: In function `qedi_ep_connect':
qedi_iscsi.c:(.text.qedi_ep_connect+0x350): undefined reference to `do_not_recover'
drivers/scsi/qedi/qedi_fw.o: In function `qedi_tmf_work':
qedi_fw.c:(.text.qedi_tmf_work+0x3b4): undefined reference to `do_not_recover'
This defines the symbol as a constant in this case, as there is no way to
set it to anything other than zero without DEBUG_FS. In addition, I'm renaming
it to qedi_do_not_recover in order to put it into a driver specific namespace,
as "do_not_recover" is a really bad name for a kernel-wide global identifier
when it is used only in one driver.
Fixes: ace7f46ba5fd ("scsi: qedi: Add QLogic FastLinQ offload iSCSI driver framework.") Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Manish Rangankar <Manish.Rangankar@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Brian Maly <brian.maly@oracle.com>
The call to qedi_setup_int is not updating the return code rc yet rc
is being checked for an error. Fix this by assigning rc to the return
code from the call to qedi_setup_int.
Signed-off-by: Colin Ian King <colin.king@canonical.com> Signed-off-by: Brian Maly <brian.maly@oracle.com>
'conn_info' is malloced in qedi_iscsi_update_conn() and should be freed
before leaving from the error handling cases, otherwise it will cause
memory leak.
Fixes: ace7f46ba5fd ("scsi: qedi: Add QLogic FastLinQ offload iSCSI driver framework.") Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com> Acked-by: Manish Rangankar <Manish.Rangankar@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Brian Maly <brian.maly@oracle.com>
Although on most systems va_end is a no-op, it is good practice to use
va_end on the function return path, especially since the va_start
documenation states:
"Each invocation of va_start() must be matched by a corresponding
invocation of va_end() in the same function."
Found with static analysis by CoverityScan, CIDs 1389477-1389479
Signed-off-by: Colin Ian King <colin.king@canonical.com> Acked-by: Manish Rangankar <manish.rangankar@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Brian Maly <brian.maly@oracle.com>
Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Cc: QLogic-Storage-Upstream@cavium.com Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Brian Maly <brian.maly@oracle.com>
log a message when we enter this situation:
1) we already allocated the max number of available grants from hypervisor
and
2) we still need more (but the request fails because of 1)).
Sometimes the lack of grants causes IO hangs in xen_blkfront devices.
Adding this log would help debuging.
Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>
We call skb_cow_data, which is good anyway to ensure we can actually
modify the skb as such (another error from prior). Now that we have the
number of fragments required, we can safely allocate exactly that amount
of memory.
Fixes: c09440f7dcb3 ("macsec: introduce IEEE 802.1AE driver") Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> Acked-by: Sabrina Dubroca <sd@queasysnail.net> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 5294b83086cc1c35b4efeca03644cf9d12282e5b) Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
drivers/net/macsec.c
While this may appear as a humdrum one line change, it's actually quite
important. An sk_buff stores data in three places:
1. A linear chunk of allocated memory in skb->data. This is the easiest
one to work with, but it precludes using scatterdata since the memory
must be linear.
2. The array skb_shinfo(skb)->frags, which is of maximum length
MAX_SKB_FRAGS. This is nice for scattergather, since these fragments
can point to different pages.
3. skb_shinfo(skb)->frag_list, which is a pointer to another sk_buff,
which in turn can have data in either (1) or (2).
The first two are rather easy to deal with, since they're of a fixed
maximum length, while the third one is not, since there can be
potentially limitless chains of fragments. Fortunately dealing with
frag_list is opt-in for drivers, so drivers don't actually have to deal
with this mess. For whatever reason, macsec decided it wanted pain, and
so it explicitly specified NETIF_F_FRAGLIST.
Because dealing with (1), (2), and (3) is insane, most users of sk_buff
doing any sort of crypto or paging operation calls a convenient function
called skb_to_sgvec (which happens to be recursive if (3) is in use!).
This takes a sk_buff as input, and writes into its output pointer an
array of scattergather list items. Sometimes people like to declare a
fixed size scattergather list on the stack; othertimes people like to
allocate a fixed size scattergather list on the heap. However, if you're
doing it in a fixed-size fashion, you really shouldn't be using
NETIF_F_FRAGLIST too (unless you're also ensuring the sk_buff and its
frag_list children arent't shared and then you check the number of
fragments in total required.)
Specifying MAX_SKB_FRAGS + 1 is the right answer usually, but not if you're
using NETIF_F_FRAGLIST, in which case the call to skb_to_sgvec will
overflow the heap, and disaster ensues.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> Cc: stable@vger.kernel.org Cc: security@kernel.org Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 4d6fa57b4dab0d77f4d8e9d9c73d1e63f6fe8fee) Signed-off-by: Brian Maly <brian.maly@oracle.com>
Conflicts:
drivers/net/macsec.c
A client can append random data to the end of an NFSv2 or NFSv3 RPC call
without our complaining; we'll just stop parsing at the end of the
expected data and ignore the rest.
Encoded arguments and replies are stored together in an array of pages,
and if a call is too large it could leave inadequate space for the
reply. This is normally OK because NFS RPC's typically have either
short arguments and long replies (like READ) or long arguments and short
replies (like WRITE). But a client that sends an incorrectly long reply
can violate those assumptions. This was observed to cause crashes.
Also, several operations increment rq_next_page in the decode routine
before checking the argument size, which can leave rq_next_page pointing
well past the end of the page array, causing trouble later in
svc_free_pages.
So, following a suggestion from Neil Brown, add a central check to
enforce our expectation that no NFSv2/v3 call has both a large call and
a large reply.
As followup we may also want to rewrite the encoding routines to check
more carefully that they aren't running off the end of the page array.
We may also consider rejecting calls that have any extra garbage
appended. That would be safer, and within our rights by spec, but given
the age of our server and the NFS protocol, and the fact that we've
never enforced this before, we may need to balance that against the
possibility of breaking some oddball client.
Reported-by: Tuomas Haanpää <thaan@synopsys.com> Reported-by: Ari Kauppi <ari@synopsys.com> Cc: stable@vger.kernel.org Reviewed-by: NeilBrown <neilb@suse.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
(cherry picked from commit e6838a29ecb484c97e4efef9429643b9851fba6e) Signed-off-by: Brian Maly <brian.maly@oracle.com>
Paolo Abeni [Thu, 27 Apr 2017 17:29:34 +0000 (19:29 +0200)]
bonding: avoid defaulting hard_header_len to ETH_HLEN on slave removal
On slave list updates, the bonding driver computes its hard_header_len
as the maximum of all enslaved devices's hard_header_len.
If the slave list is empty, e.g. on last enslaved device removal,
ETH_HLEN is used.
Since the bonding header_ops are set only when the first enslaved
device is attached, the above can lead to header_ops->create()
being called with the wrong skb headroom in place.
If bond0 is configured on top of ipoib devices, with the
following commands:
ifup bond0
for slave in $BOND_SLAVES_LIST; do
ip link set dev $slave nomaster
done
ping -c 1 <ip on bond0 subnet>
we will obtain a skb_under_panic() with a similar call trace:
skb_push+0x3d/0x40
push_pseudo_header+0x17/0x30 [ib_ipoib]
ipoib_hard_header+0x4e/0x80 [ib_ipoib]
arp_create+0x12f/0x220
arp_send_dst.part.19+0x28/0x50
arp_solicit+0x115/0x290
neigh_probe+0x4d/0x70
__neigh_event_send+0xa7/0x230
neigh_resolve_output+0x12e/0x1c0
ip_finish_output2+0x14b/0x390
ip_finish_output+0x136/0x1e0
ip_output+0x76/0xe0
ip_local_out+0x35/0x40
ip_send_skb+0x19/0x40
ip_push_pending_frames+0x33/0x40
raw_sendmsg+0x7d3/0xb50
inet_sendmsg+0x31/0xb0
sock_sendmsg+0x38/0x50
SYSC_sendto+0x102/0x190
SyS_sendto+0xe/0x10
do_syscall_64+0x67/0x180
entry_SYSCALL64_slow_path+0x25/0x25
This change addresses the issue avoiding updating the bonding device
hard_header_len when the slaves list become empty, forbidding to
shrink it below the value used by header_ops->create().
The bug is there since commit 54ef31371407 ("[PATCH] bonding: Handle large
hard_header_len") but the panic can be triggered only since
commit fc791b633515 ("IB/ipoib: move back IB LL address into the hard
header").
Reported-by: Norbert P <noe@physik.uzh.ch> Fixes: 54ef31371407 ("[PATCH] bonding: Handle large hard_header_len") Fixes: fc791b633515 ("IB/ipoib: move back IB LL address into the hard header") Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 19cdead3e2ef8ed765c5d1ce48057ca9d97b5094)
Rama Nichanamatlu [Tue, 27 Jun 2017 12:34:16 +0000 (05:34 -0700)]
[PATCH] RDS: Print failed rdma op details if failure is remote access
Improves diagnosability when RDMA op fails allowing this print to be
matched with prints on the responder side; which prints RDMA keys
which are prematurely purged by the owning process.
Rama Nichanamatlu [Tue, 27 Jun 2017 12:22:46 +0000 (05:22 -0700)]
[PATCH] RDS: When RDS socket is closed, print unreleased MR's
Improves diagnosability when the requester RDMA operation fails with
remote access error. This commit prints RDMA credentials which are
prematurely purged by owning process.
Qing Huang [Thu, 18 May 2017 23:33:53 +0000 (16:33 -0700)]
RDMA/core: not to set page dirty bit if it's already set.
This change will optimize kernel memory deregistration operations.
__ib_umem_release() used to call set_page_dirty_lock() against every
writable page in its memory region. Its purpose is to keep data
synced between CPU and DMA device when swapping happens after mem
deregistration ops. Now we choose not to set page dirty bit if it's
already set by kernel prior to calling __ib_umem_release(). This
reduces memory deregistration time by half or even more when we ran
application simulation test program.
8K + 256 (8448B) is an important message size for the RDBMS workload. Since
Infiniband supports scatter-gather in hardware, there is no reason to
fragment each RDS message into PAGE_SIZE work requests. Hence, RDS fragment
sizes up to 16K has been introduced.
Fixes: 23f90cccfba4 ("RDS: fix the sg allocation based on actual msg sz")
Previous behavior was allocating a contiguous memory buffer, corresponding
to the size of the RDS message. Although this was functional correct, it
introduced hard pressure on the memory allocation system, which was
not needed.
This commit fixes the drawback introduced by only allocating
the buffer according to RDS_MAX_FRAG_SIZE.
Sagi Grimberg [Tue, 6 Oct 2015 16:52:37 +0000 (19:52 +0300)]
xprtrdma: Don't require LOCAL_DMA_LKEY support for fastreg
There is no need to require LOCAL_DMA_LKEY support as the
PD allocation makes sure that there is a local_dma_lkey. Also
correctly set a return value in error path.
This caused a NULL pointer dereference in mlx5 which removed
the support for LOCAL_DMA_LKEY.
Fixes: bb6c96d72879 ("xprtrdma: Replace global lkey with lkey local to PD") Signed-off-by: Sagi Grimberg <sagig@mellanox.com> Reviewed-by: Chuck Lever <chuck.lever@oracle.com> Acked-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Orabug: 26151481
Alexander Duyck [Mon, 30 Jan 2017 20:29:35 +0000 (12:29 -0800)]
i40e/i40evf: Add support for mapping pages with DMA attributes
This patch adds support for DMA_ATTR_SKIP_CPU_SYNC and
DMA_ATTR_WEAK_ORDERING. By enabling both of these for the Rx path we
are able to see performance improvements on architectures that implement
either one due to the fact that page mapping and unmapping only has to
sync what is actually being used instead of the entire buffer. In addition
by enabling the weak ordering attribute enables a performance improvement
for architectures that can associate a memory ordering with a DMA buffer
such as Sparc.
Change-ID: If176824e8231c5b24b8a5d55b339a6026738fc75 Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Orabug: 26396243
Local modifications to account for dma_attr data type difference.
(cherry picked from commit 59605bc09630c2b577858c371edf89c099b5f925) Signed-off-by: Jack Vogel <jack.vogel@oracle.com> Reviewed-by: Shannon Nelson <shannon.nelson@oracle.com>
Christoph Hellwig [Fri, 23 Jun 2017 17:41:41 +0000 (10:41 -0700)]
block: defer timeouts to a workqueue
Timer context is not very useful for drivers to perform any meaningful abort
action from. So instead of calling the driver from this useless context
defer it to a workqueue as soon as possible.
Note that while a delayed_work item would seem the right thing here I didn't
dare to use it due to the magic in blk_add_timer that pokes deep into timer
internals. But maybe this encourages Tejun to add a sensible API for that to
the workqueue API and we'll all be fine in the end :)
Contains a major update from Keith Bush:
"This patch removes synchronizing the timeout work so that the timer can
start a freeze on its own queue. The timer enters the queue, so timer
context can only start a freeze, but not wait for frozen."
Signed-off-by: Christoph Hellwig <hch@lst.de> Acked-by: Keith Busch <keith.busch@intel.com> Signed-off-by: Jens Axboe <axboe@fb.com>
(cherry picked from commit 287922eb0b186e2a5bf54fdd04b734c25c90035c)
Rob Gardner [Fri, 9 Jun 2017 04:36:24 +0000 (00:36 -0400)]
sparc64: Set valid bytes of misaligned no-fault loads
If a misaligned no-fault load (ldm* from ASI 0x82, primary no fault)
crosses a page boundary, and one of the pages causes an MMU miss
that cannot be resolved, then the kernel must load bytes from the
valid page into the high (or low) bytes of the destination register,
and must load zeros into the low (or high) bytes of the register.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com> Reviewed-by: Steve Sistare steven.sistare@oracle.com Reviewed-by: Anthony Yznaga <anthony.yznaga@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Babu Moger [Wed, 21 Jun 2017 23:22:09 +0000 (17:22 -0600)]
fs/fuse: Fix for correct number of numa nodes
When fuse filesystem is mounted it sets up data structure
for all the available numa nodes(with -o numa). However,
it uses nr_node_ids which is set to MAX_NUMNODES(16). This
causes following panic when kmalloc_node is called.
Pavel Tatashin [Thu, 15 Jun 2017 14:40:59 +0000 (10:40 -0400)]
sparc64: broken %tick frequency on spitfire cpus
After early boot time stamps project the %tick frequency is detected
incorrectly on spittfire cpus.
We must use cpuid of boot cpu to find corresponding cpu node in OpenBoot,
and extract clock-frequency property from there.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit eea9833453bd39e2f35325abb985d00486c8aa69) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Thu, 15 Jun 2017 14:40:58 +0000 (10:40 -0400)]
sparc64: use prom interface to get %stick frequency
We initialize time early, we must use prom interface instead of open
firmware driver, which is not yet initialized.
Also, use prom_getintdefault() instead of prom_getint() to be compatible
with the code before early boot timestamps project.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit fca4afe400cb68fe5a7f0a97fb1ba5cfdcb81675) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Mon, 12 Jun 2017 20:41:48 +0000 (16:41 -0400)]
sparc64: optimize functions that access tick
Replace read tick function pointers with the new hot-patched get_tick().
This optimizes the performance of functions such as: sched_clock()
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit eae3fc9871111e9bbc77dad5481a3e805e02ac46) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Mon, 12 Jun 2017 20:41:47 +0000 (16:41 -0400)]
sparc64: add hot-patched and inlined get_tick()
Add the new get_tick() function that is hot-patched during boot based on
processor we are booting on.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit 4929c83a6ce6584cb64381bf1407c487f67d588a) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Mon, 12 Jun 2017 20:41:46 +0000 (16:41 -0400)]
sparc64: initialize time early
In Linux it is possible to configure printk() to output timestamp next to
every line. This is very useful to determine the slow parts of the boot
process, and also to avoid regressions, as boot time is visiable to
everyone.
Also, there are scripts that change these time stamps to intervals.
However, on larger machines these timestamps start appearing many seconds,
and even minutes into the boot process. This patch gets stick-frequency
property early from OpenBoot, and uses its value to initialize time stamps
before the first printk() messages are printed.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Shannon Nelson <shannon.nelson@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit 83e8eb99d908da78e6eff7dd141f26626fe01d12) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Mon, 12 Jun 2017 20:41:45 +0000 (16:41 -0400)]
sparc64: improve modularity tick options
This patch prepares the code for early boot time stamps by making it more
modular.
- init_tick_ops() to initialize struct sparc64_tick_ops
- new sparc64_tick_ops operation get_frequency() which returns a
frequency
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit 89108c3423e8047cd0da73182ea09b9da190b57e) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Mon, 12 Jun 2017 20:41:44 +0000 (16:41 -0400)]
sparc64: optimize loads in clock_sched()
In clock sched we now have three loads:
- Function pointer
- quotient for multiplication
- offset
However, it is possible to improve performance substantially, by
guaranteeing that all three loads are from the same cacheline.
By moving these three values first in sparc64_tick_ops, and by having
tick_operations 64-byte aligned we guarantee this.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Shannon Nelson <shannon.nelson@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit 178bf2b9a20e866677bbca5cb521b09a8498c1d7) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Mon, 12 Jun 2017 20:41:43 +0000 (16:41 -0400)]
sparc64: show time stamps from zero
On most platforms, time is shown from the beginning of boot. This patch is
adding offset to sched_clock() for SPARC, to also show time from 0.
This means we will have one more load, but we saved one in an ealier patch.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit b5dd4d807f0fe7da67c5cc67b2ec681b60e4994b) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Shannon Nelson <shannon.nelson@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit b8a83fcb78c859b99807af4c8b0ab09f0f827a40) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Mon, 12 Jun 2017 20:41:41 +0000 (16:41 -0400)]
sparc64: remove trailing white spaces
A few changes that were reported by checkpatch, removed all trailing white
spaces in these two files.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Shannon Nelson <shannon.nelson@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Orabug: 24401250
Orabug: 25637776
(cherry picked from commit 68a792174d7f67c7d2108bf1cc55ab8a63fc4678) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Wed, 31 May 2017 15:25:25 +0000 (11:25 -0400)]
sparc64: delete old wrap code
The old method that is using xcall and softint to get new context id is
deleted, as it is replaced by a method of using per_cpu_secondary_mm
without xcall to perform the context wrap.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 0197e41ce70511dc3b71f7fefa1a676e2b5cd60b) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Wed, 31 May 2017 15:25:24 +0000 (11:25 -0400)]
sparc64: new context wrap
The current wrap implementation has a race issue: it is called outside of
the ctx_alloc_lock, and also does not wait for all CPUs to complete the
wrap. This means that a thread can get a new context with a new version
and another thread might still be running with the same context. The
problem is especially severe on CPUs with shared TLBs, like sun4v. I used
the following test to very quickly reproduce the problem:
- start over 8K processes (must be more than context IDs)
- write and read values at a memory location in every process.
Very quickly memory corruptions start happening, and what we read back
does not equal what we wrote.
Several approaches were explored before settling on this one:
Approach 1:
Move smp_new_mmu_context_version() inside ctx_alloc_lock, and wait for
every process to complete the wrap. (Note: every CPU must WAIT before
leaving smp_new_mmu_context_version_client() until every one arrives).
This approach ends up with deadlocks, as some threads own locks which other
threads are waiting for, and they never receive softint until these threads
exit smp_new_mmu_context_version_client(). Since we do not allow the exit,
deadlock happens.
Approach 2:
Handle wrap right during mondo interrupt. Use etrap/rtrap to enter into
into C code, and issue new versions to every CPU.
This approach adds some overhead to runtime: in switch_mm() we must add
some checks to make sure that versions have not changed due to wrap while
we were loading the new secondary context. (could be protected by PSTATE_IE
but that degrades performance as on M7 and older CPUs as it takes 50 cycles
for each access). Also, we still need a global per-cpu array of MMs to know
where we need to load new contexts, otherwise we can change context to a
thread that is going way (if we received mondo between switch_mm() and
switch_to() time). Finally, there are some issues with window registers in
rtrap() when context IDs are changed during CPU mondo time.
The approach in this patch is the simplest and has almost no impact on
runtime. We use the array with mm's where last secondary contexts were
loaded onto CPUs and bump their versions to the new generation without
changing context IDs. If a new process comes in to get a context ID, it
will go through get_new_mmu_context() because of version mismatch. But the
running processes do not need to be interrupted. And wrap is quicker as we
do not need to xcall and wait for everyone to receive and complete wrap.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit a0582f26ec9dfd5360ea2f35dd9a1b026f8adda0) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 7a5b4bbf49fe86ce77488a70c5dccfe2d50d7a2d) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Wed, 31 May 2017 15:25:22 +0000 (11:25 -0400)]
sparc64: redefine first version
CTX_FIRST_VERSION defines the first context version, but also it defines
first context. This patch redefines it to only include the first context
version.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit c4415235b2be0cc791572e8e7f7466ab8f73a2bf) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Wed, 31 May 2017 15:25:21 +0000 (11:25 -0400)]
sparc64: combine activate_mm and switch_mm
The only difference between these two functions is that in activate_mm we
unconditionally flush context. However, there is no need to keep this
difference after fixing a bug where cpumask was not reset on a wrap. So, in
this patch we combine these.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 14d0334c6748ff2aedb3f2f7fdc51ee90a9b54e7) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
Pavel Tatashin [Wed, 31 May 2017 15:25:20 +0000 (11:25 -0400)]
sparc64: reset mm cpumask after wrap
After a wrap (getting a new context version) a process must get a new
context id, which means that we would need to flush the context id from
the TLB before running for the first time with this ID on every CPU. But,
we use mm_cpumask to determine if this process has been running on this CPU
before, and this mask is not reset after a wrap. So, there are two possible
fixes for this issue:
1. Clear mm cpumask whenever mm gets a new context id
2. Unconditionally flush context every time process is running on a CPU
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 588974857359861891f478a070b1dc7ae04a3880) Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: Allen Pais <allen.pais@oracle.com>
projects / users / jedix / linux-maple.git / log