| commit | 6a581e93981e8838c85e407303186faf937830d3 | |
| author | Dhananjay Phadke <dhananjay@netxen.com> | |
| Sat, 5 Sep 2009 17:43:08 +0000 (5 17:43 +0000) | ||
| committer | David S. Miller <davem@davemloft.net> | |
| Mon, 7 Sep 2009 08:53:06 +0000 (7 01:53 -0700) | ||
| tree | 3e2f083971fc33d4a3e88e7cc37e346f775d7168 | treesnapshot (tar.gz zip) |
| parent | db4cfd8a6149e778befb2ff6e6f91cdc6394cbe6 | commitdiff |
netxen: firmware hang detection
Implement state machine to detect firmware hung state
and recover. Since firmware will be shared by all PCI
functions that have different class drivers (NIC or
FCOE or iSCSI), explicit hardware based serialization
is required for initializing firmware.
o Used global scratchpad register to maintain device
reference count. Every probed pci function adds to
ref count.
o Implement timer (delayed work) for each pci func
that checks firmware heartbit every 5 sec and detaches
itself if firmware is dead. Last detaching function
reloads firmware. Other functions wait for firmware
init, and re-attach themselves.
Heartbit is not supported by NX2031 firmware.
Signed-off-by: Amit Kumar Salecha <amit@netxen.com>
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement state machine to detect firmware hung state
and recover. Since firmware will be shared by all PCI
functions that have different class drivers (NIC or
FCOE or iSCSI), explicit hardware based serialization
is required for initializing firmware.
o Used global scratchpad register to maintain device
reference count. Every probed pci function adds to
ref count.
o Implement timer (delayed work) for each pci func
that checks firmware heartbit every 5 sec and detaches
itself if firmware is dead. Last detaching function
reloads firmware. Other functions wait for firmware
init, and re-attach themselves.
Heartbit is not supported by NX2031 firmware.
Signed-off-by: Amit Kumar Salecha <amit@netxen.com>
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>