In its current state, the driver will handle backing device
login in a loop for a certain number of retries while the
device returns a partial success, indicating that the driver
may need to try again using a smaller number of resources.
The variable it checks to continue retrying may change
over the course of operations, resulting in reallocation
of resources but exits without sending the login attempt.
Guard against this by introducing a boolean variable that
will retain the state indicating that the driver needs to
reattempt login with backing device firmware.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce a recovery hard reset to handle reset failure as a result of
change of device context following a transport event, such as a
backing device failover or partition migration. These operations reset
the device context to its initial state. If this occurs during a reset,
any initialization commands are likely to fail with an invalid state
error as backing device firmware requests reinitialization.
When this happens, make one more attempt by performing a hard reset,
which frees any resources currently allocated and performs device
initialization. If a transport event occurs during a device reset, a
flag is set which will trigger a new hard reset following the
completionof the current reset event.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set device resetting state at the earliest possible point: as soon as a
reset is successfully scheduled. The reset state is toggled off when
all resets have been processed to completion.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of having one initialization routine for all cases, create
a separate, simpler function for standard initialization, such as during
device probe. Use the original initialization function to handle
device reset scenarios. The goal of this patch is to avoid having
a single, cluttered init function to handle all possible
scenarios.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If setting the link state is not successful, print a warning
with the resulting return code and return it to be handled
by the caller.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If device init is interrupted by a failover, set the init return
code so that it can be checked and handled appropriately by the
init routine.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Check whether CRQ command is successful before awaiting a response
from the management partition. If the command was not successful, the
driver may hang waiting for a response that will never come.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce an "active" state for a IBM vNIC Command-Response Queue. A CRQ
is considered active once it has initialized or linked with its partner by
sending an initialization request and getting a successful response back
from the management partition. Until this has happened, do not allow CRQ
commands to be sent other than the initialization request.
This change will avoid a protocol error in case of a device transport
event occurring during a initialization. When the driver receives a
transport event notification indicating that the backing hardware
has changed and needs reinitialization, any further commands other
than the initialization handshake with the VIOS management partition
will result in an invalid state error. Instead of sending a command
that will be returned with an error, print a warning and return an
error that will be handled by the caller.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set adapter NAPI state as disabled if they are removed. This will allow
them to be enabled again if reallocated in case of a hard reset.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When enabling the sub-CRQ IRQ a previous update sent a H_EOI prior
to the enablement to clear any pending interrupts that may be present
across a partition migration. This fixed a firmware bug where a
migration could erroneously indicate that a H_EOI was pending.
The H_EOI should only be sent when enabling during a mobility
event though. Doing so at other time could wrong and can produce
extra driver output when IRQs are enabled when doing TX completion.
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move initialization of statistics buffers from ibmvnic_init function
into ibmvnic_probe. In the current state, ibmvnic_init will be called
again during a device reset, resulting in the allocation of new
buffers without freeing the old ones.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is not necessary to disable interrupt lines here during a reset
to handle a non-fatal firmware error. Move that call within the code
block that handles the other cases that do require interrupts to be
disabled and re-enabled.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the firmware map fails for whatever reason, remember to free
up the memory after.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Avoid using value stored in the login response buffer when
cleaning TX and RX buffer pools since these could be inconsistent
depending on the device state. Instead use the field in the driver's
private data that tracks the number of active pools.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to a firmware bug, the hypervisor can send an interrupt to a
transmit or receive queue just prior to a partition migration, not
allowing the device enough time to handle it and send an EOI. When
the partition migrates, the interrupt is lost but an "EOI-pending"
flag for the interrupt line is still set in firmware. No further
interrupts will be sent until that flag is cleared, effectively
freezing that queue. To workaround this, the driver will disable the
hardware interrupt and send an H_EOI signal prior to re-enabling it.
This will flush the pending EOI and allow the driver to continue
operation.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When attempting to change the driver parameters, such as the MTU
value or number of queues, do not call netdev_notify_peers().
Doing so will deadlock on the rtnl_lock.
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a bug in handling the possible return codes from sending the
login CRQ. The current code treats any non-success return value,
minus failure to send the crq and a timeout waiting for a login response,
as a need to re-send the login CRQ. This can put the drive in an
infinite loop of trying to login when getting return values other
that a partial success such as a return code of aborted. For these
scenarios the login will not ever succeed at this point and the
driver would need to be reset again.
To resolve this loop trying to login is updated to only retry the
login if the driver gets a return code of a partial success. Other
return codes are treated as an error and the driver returns an error
from ibmvnic_login().
To avoid infinite looping in the partial success return cases, the
number of retries is capped at the maximum number of supported
queues. This value was chosen because the driver does a renegotiation
of capabilities which sets the number of queues possible and allows
the driver to attempt a login for possible value for the number
of queues supported.
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The "name" field of struct vnic_login_client_data is a char array of
undefined length. This should be written as "char name[]" so the compiler
can make better decisions about the field (for example, not assuming
it's a single character). This was noticed while trying to tighten the
CONFIG_FORTIFY_SOURCE checking.
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When resetting the ibmvnic driver after a partition migration occurs
there is no requirement to do a reset of the main CRQ. The current
driver code does the required re-enable of the main CRQ, then does
a reset of the main CRQ later.
What we should be doing for a driver reset after a migration is to
re-enable the main CRQ, release all the sub-CRQs, and then allocate
new sub-CRQs after capability negotiation.
This patch updates the handling of mobility resets to do the proper
work and not reset the main CRQ. To do this the initialization/reset
of the main CRQ had to be moved out of the ibmvnic_init routine
and in to the ibmvnic_probe and do_reset routines.
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a failover case for a non-redundant pseries VNIC
configuration that was not being handled properly. The current
implementation assumes that the driver will always have a redandant
device to communicate with following a failover notification. There
are cases, however, when a non-redundant configuration can receive
a failover request. If that happens, the driver should wait until
it receives a signal that the device is ready for operation.
The driver is agnostic of its backing hardware configuration,
so this fix necessarily affects all device failover management.
The driver needs to wait until it receives a signal that the device
is ready for resetting. A flag is introduced to track this intermediary
state where the driver is waiting for an active device.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In some cases, if the driver is waiting for a reset following
a device parameter change, failure to schedule a reset can result
in a hang since a completion signal is never sent.
If the device configuration is being altered by a tool such
as ethtool or ifconfig, it could cause the console to hang
if the reset request does not get scheduled. Add some additional
error handling code to exit the wait_for_completion if there is
one in progress.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The counter that tracks used TX descriptors pending completion
needs to be zeroed as part of a device reset. This change fixes
a bug causing transmit queues to be stopped unnecessarily and in
some cases a transmit queue stall and timeout reset. If the counter
is not reset, the remaining descriptors will not be "removed",
effectively reducing queue capacity. If the queue is over half full,
it will cause the queue to stall if stopped.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix some mistakes caught by the DMA debugger. The first change
fixes a unnecessary unmap that should have been removed in an
earlier update. The next hunk fixes another bad unmap by zeroing
the bit checked to determine that an unmap is needed. The final
change fixes some buffers that are unmapped with the wrong
direction specified.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the driver is closed, all the associated irqs are disabled. In the
event that the driver exits a reset in the closed state, we should be
consistent with the state we are in directly after a close. So before we
exit the reset routine, all irqs should be disabled as well. This will
prevent the irqs from being enabled twice in this case and reporting a
number of noisy warning traces.
Signed-off-by: John Allen <jallen@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prefer the direct use of octal for permissions.
Done with checkpatch -f --types=SYMBOLIC_PERMS --fix-inplace
and some typing.
Miscellanea:
o Whitespace neatening around these conversions.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is an && vs || typo here, which potentially leads to a NULL
dereference.
Fixes: e9e1e97884 ("ibmvnic: Update TX pool cleaning routine")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Finally, remove the TSO-specific fields in the TX pool
strcutures. These are no longer needed with the introduction
of separate buffer pools for TSO transmissions.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Update routine that cleans up any outstanding transmits that
have not received completions when the device needs to close.
Introduces a helper function that cleans one TX pool to make
code more readable.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Improve TX pool buffer accounting to prevent the producer
index from overruning the consumer. First, set the next free
index to an invalid value if it is in use. If next buffer
to be consumed is in use, drop the packet.
Finally, if the transmit fails for some other reason, roll
back the consumer index and set the free map entry to its original
value. This should also be done if the DMA map fails.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Update TX and TX completion routines to account for TX pool
restructuring. TX routine first chooses the pool depending
on whether a packet is GSO or not, then uses it accordingly.
For the completion routine to know which pool it needs to use,
set the most significant bit of the correlator index to one
if the packet uses the TSO pool. On completion, unset the bit
and use the correlator index to release the buffer pool entry.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce function that initializes one TX pool. Use that to
create each pool entry in both the standard TX pool and TSO
pool arrays.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce function that frees one TX pool. Use that to release
each pool in both the standard TX pool and TSO pool arrays.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Update TX pool reset routine to accommodate new TSO pool array. Introduce
a function that resets one TX pool, and use that function to initialize
each pool in both pool arrays.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove some unused fields in the structure and include values
describing the individual buffer size and number of buffers in
a TX pool. This allows us to use these fields for TX pool buffer
accounting as opposed to using hard coded values. Include a new
pool array for TSO transmissions.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The case in which we handle a reset from the state where the device is
closed seems to be bugged for all types of reset. For most types of reset
we currently exit the reset routine correctly, but don't set the state to
indicate that we are back in the "closed" state. For some specific cases,
we don't exit the reset routine at all and resetting will cause a closed
device to be opened.
This patch fixes the problem by unconditionally checking the reset_state
and correctly setting the adapter state before returning.
Signed-off-by: John Allen <jallen@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sorry, one of the patches I sent in an earlier series
has some dumb mistakes. One was that I had changed the
parameter for the errata workaround function but forgot
to make that change in the code that called it.
The second mistake was a forgotten return value at the end
of the function in case the workaround was not needed.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TSO packets with one segment or with an MSS less than 224 can
cause errors on some backing devices, so disable GSO in those cases.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some backing devices cannot handle small packets well,
so pad any small packets to avoid that. It was recommended
that the VNIC driver should not send packets smaller than the
minimum MTU value provided by firmware, so pad small packets
to be at least that long.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The extra four bytes of a VLAN packet was throwing off
TX buffer entry values used by the driver. Account for those
bytes when in buffer size and buffer entry calculations
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If a VLAN tag is present in the Ethernet header, account
for that when providing the L2 header to firmware.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During a device failover or partition migration reset, it is not
necessary to disable the backing adapter since it should not be
running yet and its Command-Response Queue is closed. Sending
device commands during this time could result in an error or
timeout disrupting the reset process. In these cases, just halt
transmissions, clean up resources, and continue with reset.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce a function to halt network operations and clean up any
unused or outstanding socket buffers. Then, during device close,
disable backing adapter before halting all queues and performing
cleanup. This ensures all backing device operations will be
stopped before the driver cleans up shared resources.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove some dead code now that RX pools are being cleaned. This
was included to wait until any pending RX queue interrupts are
processed, but NAPI polling should be disabled by this point.
Another minor change is to use the net device parameter for any
print functions instead of accessing it from the adapter structure.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If a device reset fails for some reason, TX and RX queue resources
could be released. If a user attempts to open the device in this scenario,
it may result in a kernel panic as the driver tries to access this
memory. To fix this, include a check before device login that TX/RX
queues are still there before enabling the device. In addition, return a
value that can be checked in case of any errors to avoid waiting for a
completion that will never come.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It's not necessary to report each time a queue is stopped and restarted
as an informational message. Change that to be a debug message so that
it can be observed if needed but not printed by default.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the driver releases resources after a failed reset or some other
error, the driver might attempt to clean up and free memory that
isn't there anymore. Include some additional checks that RX/TX queues
along with their associated structures are still there before cleaning.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, buffers holding individual queue statistics are allocated
when the device is opened. If an ibmvnic interface is hotplugged or
initialized but never opened, an attempt to get statistics with
ethtool will result in a kernel panic.
Since the driver allocates a constant number, the maximum supported
queues, of buffers, these can be allocated during device probe and
freed when the device is hot-unplugged or the module is removed.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sorry, the previous change introduced a race condition between
transmit completion processing and tracking TX descriptors. If a
completion is received before the number of descriptors is logged,
the number of descriptors will be add but not removed. After enough
times, this could halt the transmit queue forever.
Log the number of descriptors used by a transmit before sending.
I stress tested the fix on two different systems running over the
weekend without any issues.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
