sata_sil24: freeze on non-dev errors reported via CERR

CERR reports errors detected during executing a command.  This doesn't
mean the error is tied to the command and can be recovered by just
issuing it again.  Many of the errors are fatal port-wide connditions
including HSM violation, host bus error and ATA bus error and require
freezing and port reset.

The freezing part wasn't implemented previously.  This used to be okay
because port resets were scheduled anyway and EH eventually resets and
recovers the port.  With PMP support added, this is no longer true.
The error condition and recover actions are attributed to the fan-out
port and the host port condition isn't properly recovered leading to
EH failures.

This patch makes CERR errors which require resets to freeze the port.
This will force host port reset and proper recovery.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Andrew Ryder <tireman@shaw.ca>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

diff --git a/drivers/ata/sata_sil24.c b/drivers/ata/sata_sil24.c
index d9c8b32b483cd4f386e10874eda8bfebee693e90..864c1c1b851138805c9aa11968d8a1567b7ab305 100644 (file)
--- a/drivers/ata/sata_sil24.c
+++ b/drivers/ata/sata_sil24.c
@@ -1094,10 +1094,13 @@ static void sil24_error_intr(struct ata_port *ap)
                if (ci && ci->desc) {
                        err_mask |= ci->err_mask;
                        action |= ci->action;
+                       if (action & ATA_EH_RESET_MASK)
+                               freeze = 1;
                        ata_ehi_push_desc(ehi, "%s", ci->desc);
                } else {
                        err_mask |= AC_ERR_OTHER;
                        action |= ATA_EH_SOFTRESET;
+                       freeze = 1;
                        ata_ehi_push_desc(ehi, "unknown command error %d",
                                          cerr);
                }