This patch fixes the following crash seen when MIC reset is invoked in
RESET_FAILED state due to device_del being called a second time on an
already deleted device:
[<ffffffff813b2295>] device_del+0x45/0x1d0
[<ffffffff813b243e>] device_unregister+0x1e/0x60
[<ffffffffa040f1c2>] scif_unregister_device+0x12/0x20 [scif_bus]
[<ffffffffa042f75a>] cosm_stop+0xaa/0xe0 [mic_cosm]
[<ffffffffa042f844>] cosm_reset_trigger_work+0x14/0x20 [mic_cosm]
The fix consists in realizing that because cosm_reset changes the
state to MIC_RESETTING, cosm_stop needs the previous state, before it
changed to MIC_RESETTING, to decide whether a hw_ops->stop had
previously been issued. This is now provided in a new cosm_device
member cdev->prev_state.
Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The COSM driver communicates with the MIC cards over SCIF. A SCIF
"server" listens for incoming connections from "client" MIC cards as
they boot. After the connection is accepted a separate work item is
scheduled for each MIC card. This work item normally stays blocked in
scif_poll but wakes up to process messages from the card.
The SCIF connection between the host and card COSM components is used
to (a) send the command to shut down the card (b) receive shutdown
status back from the card upon completion of shutdown (c) receive
periodic heartbeat messages to detect card crashes (d) send host time
to the card to enable the card to sync its time to the host.
Reviewed-by: Nikhil Rao <nikhil.rao@intel.com>
Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The COSM driver allows boot, shutdown and reset of Intel MIC devices
via sysfs. This functionality was previously present in the Intel MIC
host driver but has now been taken out into a separate driver so that
it can be shared between multiple generations of Intel MIC products.
The sysfs kernel ABI used by the COSM driver is the same as that
defined originally for the MIC host driver in
Documentation/ABI/testing/sysfs-class-mic.txt.
The COSM driver also contains support for dumping the MIC card log_buf
and doing a "force reset" for the card via debugfs. The OSPM support
present in the MIC host driver has now largely been moved to user
space and only a small required OSPM functionality is now present in
the driver.
Reviewed-by: Nikhil Rao <nikhil.rao@intel.com>
Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com>
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
