]> www.infradead.org Git - nvme.git/commit
can: mcp251xfd: fix infinite loop when xmit fails
authorVitor Soares <vitor.soares@toradex.com>
Fri, 17 May 2024 13:43:55 +0000 (14:43 +0100)
committerMarc Kleine-Budde <mkl@pengutronix.de>
Fri, 21 Jun 2024 08:50:20 +0000 (10:50 +0200)
commitd8fb63e46c884c898a38f061c2330f7729e75510
treeb255e153d1a2772b7a15129abaf8f146eb3ffe60
parent0d34d8163fd87978a6abd792e2d8ad849f4c3d57
can: mcp251xfd: fix infinite loop when xmit fails

When the mcp251xfd_start_xmit() function fails, the driver stops
processing messages, and the interrupt routine does not return,
running indefinitely even after killing the running application.

Error messages:
[  441.298819] mcp251xfd spi2.0 can0: ERROR in mcp251xfd_start_xmit: -16
[  441.306498] mcp251xfd spi2.0 can0: Transmit Event FIFO buffer not empty. (seq=0x000017c7, tef_tail=0x000017cf, tef_head=0x000017d0, tx_head=0x000017d3).
... and repeat forever.

The issue can be triggered when multiple devices share the same SPI
interface. And there is concurrent access to the bus.

The problem occurs because tx_ring->head increments even if
mcp251xfd_start_xmit() fails. Consequently, the driver skips one TX
package while still expecting a response in
mcp251xfd_handle_tefif_one().

Resolve the issue by starting a workqueue to write the tx obj
synchronously if err = -EBUSY. In case of another error, decrement
tx_ring->head, remove skb from the echo stack, and drop the message.

Fixes: 55e5b97f003e ("can: mcp25xxfd: add driver for Microchip MCP25xxFD SPI CAN")
Cc: stable@vger.kernel.org
Signed-off-by: Vitor Soares <vitor.soares@toradex.com>
Link: https://lore.kernel.org/all/20240517134355.770777-1-ivitro@gmail.com
[mkl: use more imperative wording in patch description]
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c
drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c
drivers/net/can/spi/mcp251xfd/mcp251xfd.h