BMC/Intel-BMC/linux.git/drivers/fsi, branch dev-4.10

drivers (fsi): occ: Change errno for releasing client during transfer

2017-11-01T04:31:03+00:00

Clients were receiving the ECANCELED errno when the client is removed during a transfer. Use ENODEV instead, as ECANCELED is for AIO ops. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Reviewed-by: Christopher Bostic Signed-off-by: Andrew Jeffery

drivers (fsi): sbefifo: Add timeout for waiting for data

2017-10-30T23:33:53+00:00

During a transfer, in the event that the SBE isn't running, the driver will wait for data to appear in the FIFO forever. This doesn't work at BMC boot time if any transfers are attempted (by OCC driver probing, for example), as it will just hang the boot. So add a simple timeout mechanism when the driver reschdules waiting for data to show up in the FIFO. Tested on witherspoon by rebooting the BMC successfully with chassis power on. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Signed-off-by: Andrew Jeffery

drivers (fsi): sbefifo: Fix xfr deletion in lists

2017-10-26T20:56:13+00:00

Deleting an element of a list while iterating means that you'll access recently freed memory when getting the next list element. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Signed-off-by: Andrew Jeffery

drivers: fsi: sbefifo: Attempt reset request during probe

2017-10-18T14:01:21+00:00

If data is found in the FIFO, the driver currently immediately fails the probe. Instead, try a reset request and poll for it's completion for a while (this protocol is defined in the SBE FIFO spec, though the timeout lengths are my own invention). OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Signed-off-by: Andrew Jeffery

drivers: fsi: sbefifo: Fix memory leak

2017-10-18T13:25:15+00:00

Transfers weren't being cleaned up if they were complete but the user never finished reading. This is very common. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Signed-off-by: Andrew Jeffery

drivers: fsi: sbefifo: Fix queued xfrs race condition

2017-10-18T11:59:58+00:00

There was a race condition where: sbefifo remove() is called, a transfer is queued, a client is waiting, and a client is released at the same time. In this situation, remove() wakes up the client waiter, which destroys the client's list of transfers. This must be done since in an error case, we don't know if the transfers are deleted or not. Then, the client is released. Since it has no transfers, it fails to cancel the queued transfer. This transfer then wakes up on the timer, where it attempts to access it's client buffers, which have been freed. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Signed-off-by: Andrew Jeffery

drivers: fsi: occ: Fix errno for canceling a read operation

2017-10-13T06:13:09+00:00

Should not return -ECANCELED when the device has been removed. Instead, use -ENODEV. Signed-off-by: Edward A. James Signed-off-by: Andrew Jeffery

fsi: occ: Fix client memory management

2017-10-09T01:52:32+00:00

Potential for bad memory access in the worker function. Now fixed by using reference counters. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Acked-by: Andrew Jeffery Signed-off-by: Joel Stanley

fsi: occ: Add cancel to remove() and fix probe()

2017-10-09T01:51:43+00:00

Need some data to indicate to clients and the rest of the driver when the device is being removed, so add a cancel boolean. Fix up both the probe and remove functions to properly handle failures and prevent deadlocks. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Reviewed-by: Andrew Jeffery [joel: Use ENODEV instead of ECANCELLED, as it is for cancelled AIO operations] Signed-off-by: Joel Stanley

fsi: occ: Poll while receiving "command in progress"

2017-10-09T01:48:48+00:00

Waiting for OCC to return something other than "command in progress" should only be done in the worker function, instead of clients repeating the entire transfer when they receive "command in progress." In this way clients don't have to check for that return status. OpenBMC-Staging-Count: 1 Signed-off-by: Edward A. James Acked-by: Andrew Jeffery Signed-off-by: Joel Stanley