Throw me a few more hints, please. I already knew about error handling
being changed; what should I use for sleep_on(), and will the driver still
be able to release the ST-DMA lock on request done if queue management is
done by the midlevel?
On second thought - the sleep_on we can perhaps do away with by cutting
out the whole lock wait scheme (lock contention between SCSI, IDE and
floppy are the least of my concerns). Lock release is currently done by
the low level driver after calls to scsi_done() which seems fine. Lock
acquisition happens in queue_command() which may sleep, causing other
commands to time out. The error handling code suffers from the same
problem - the SCSI timer needs to be stopped while queueing commands and
doing error handling. Is there a way to guarantee this?
I've looked at the SCSI situation in more detail, and updated the Atari
SCSI code to cope with the changes in 2.6 (scsi_error expect different
return codes; queuecmnd called from softirq and cannot sleep). I get no
more lockups, but writing to a ZIP drive corrupts data on the drive. May
just be a broken ZIP drive, who knows. I need to try with a real SCSI
disk...
Christian, please apply this on top of my other 2.6.18 patch.
--- linux-2.6.18-m68k-ms/drivers/scsi/atari_scsi.c 2006-10-19 14:23:32.000000000 +0200
+++ linux-2.6.18-m68k/drivers/scsi/atari_scsi.c 2006-12-31 12:28:08.895670559 +0100
@@ -67,12 +67,39 @@
#include <linux/module.h>
-#define NDEBUG (0)
+#define NDEBUG_ARBITRATION 0x1
+#define NDEBUG_AUTOSENSE 0x2
+#define NDEBUG_DMA 0x4
+#define NDEBUG_HANDSHAKE 0x8
+#define NDEBUG_INFORMATION 0x10
+#define NDEBUG_INIT 0x20
+#define NDEBUG_INTR 0x40
+#define NDEBUG_LINKED 0x80
+#define NDEBUG_MAIN 0x100
+#define NDEBUG_NO_DATAOUT 0x200
+#define NDEBUG_NO_WRITE 0x400
+#define NDEBUG_PIO 0x800
+#define NDEBUG_PSEUDO_DMA 0x1000
+#define NDEBUG_QUEUES 0x2000
+#define NDEBUG_RESELECTION 0x4000
+#define NDEBUG_SELECTION 0x8000
+#define NDEBUG_USLEEP 0x10000
+#define NDEBUG_LAST_BYTE_SENT 0x20000
+#define NDEBUG_RESTART_SELECT 0x40000
+#define NDEBUG_EXTENDED 0x80000
+#define NDEBUG_C400_PREAD 0x100000
+#define NDEBUG_C400_PWRITE 0x200000
+#define NDEBUG_LISTS 0x400000
#define NDEBUG_ABORT 0x800000
#define NDEBUG_TAGS 0x1000000
#define NDEBUG_MERGING 0x2000000
+#define NDEBUG_ANY 0xFFFFFFFFUL
+
+#define NDEBUG (0)
+//#define NDEBUG (NDEBUG_MAIN)
+
#define AUTOSENSE
/* For the Atari version, use only polled IO or REAL_DMA */
#define REAL_DMA
@@ -198,7 +225,7 @@
static irqreturn_t scsi_falcon_intr( int irq, void *dummy, struct pt_regs *fp);
static void falcon_release_lock_if_possible( struct NCR5380_hostdata *
hostdata );
-static void falcon_get_lock( void );
+static int falcon_get_lock( void );
#ifdef CONFIG_ATARI_SCSI_RESET_BOOT
static void atari_scsi_reset_boot( void );
#endif
@@ -506,6 +533,15 @@
* again (but others waiting longer more probably will win).
*/
+/* MSch 20061228: in 2.6, the fairness wait appears to open a race with
+ * the IDE driver's use of the lock, resulting in scheduling_in_interrupt
+ * level BUG() messages.
+ * The low level queuecmd function now appears to be called from soft
+ * interrupt context (block queue task??) and cannot sleep on the lock
+ * anymore once IDE has stolen the lock.
+ * Can we return 'please retry later' to the block queue task or mid level??
+ * MSch 20061229: the race persists regardless ... leave it off for now.
+ */
static void
falcon_release_lock_if_possible( struct NCR5380_hostdata * hostdata )
{
@@ -529,7 +565,9 @@
}
falcon_got_lock = 0;
stdma_release();
+#if defined(FALCON_FAIRNESS_WAIT)
wake_up( &falcon_fairness_wait );
+#endif
}
local_irq_restore(flags);
@@ -550,20 +588,37 @@
* Complicated, complicated.... Sigh...
*/
-static void falcon_get_lock( void )
+/* MSch 20061229: atari_queue_command gets called from softirq context quite
+ * heavily in the 2.6 kernel series. Since atari_queue_command might need to
+ * sleep in order to grab the ST-DMA lock, I have modified falcon_get_lock to
+ * immediately return with error status if called in softirq context with the
+ * lock not currently held by the SCSI driver, and the ST-DMA locked by some
+ * other driver. atari_queue_command then returns SCSI_MLQUEUE_HOST_BUSY and
+ * prevents further commands from issueing.
+ */
+
+static int falcon_get_lock( void )
{
unsigned long flags;
- if (IS_A_TT()) return;
+ if (IS_A_TT()) return 0;
local_irq_save(flags);
+#if defined (FALCON_FAIRNESS_WAIT)
while( !in_irq() && falcon_got_lock && stdma_others_waiting() )
sleep_on( &falcon_fairness_wait );
+#endif
while (!falcon_got_lock) {
if (in_irq())
panic( "Falcon SCSI hasn't ST-DMA lock in interrupt" );
+ /* we may not sleep in soft interrupts neither, so bail out */
+ if (in_softirq() && stdma_islocked()) {
+ printk(KERN_INFO "Falcon SCSI does not hold ST-DMA lock in softirq!\n" );
+ local_irq_restore(flags);
+ return 1;
+ }
if (!falcon_trying_lock) {
falcon_trying_lock = 1;
stdma_lock(scsi_falcon_intr, NULL);
@@ -579,6 +634,8 @@
local_irq_restore(flags);
if (!falcon_got_lock)
panic("Falcon SCSI: someone stole the lock :-(\n");
+
+ return 0;
}
@@ -824,6 +881,8 @@
struct NCR5380_hostdata *hostdata =
(struct NCR5380_hostdata *)cmd->device->host->hostdata;
+ printk( "scsi%d: resetting the SCSI bus!\n", (cmd)->device->host->host_no);
+
/* For doing the reset, SCSI interrupts must be disabled first,
* since the 5380 raises its IRQ line while _RST is active and we
* can't disable interrupts completely, since we need the timer.
@@ -853,8 +912,10 @@
else {
atari_turnon_irq( IRQ_MFP_FSCSI );
}
- if ((rv & SCSI_RESET_ACTION) == SCSI_RESET_SUCCESS)
+ if (rv == SUCCESS) {
falcon_release_lock_if_possible(hostdata);
+ }
+ printk( "scsi%d: bus reset done!\n", (cmd)->device->host->host_no);
return( rv );
}
--- linux-2.6.18-m68k-ms/drivers/scsi/atari_NCR5380.c 2006-10-19 14:23:32.000000000 +0200
+++ linux-2.6.18-m68k/drivers/scsi/atari_NCR5380.c 2006-12-31 12:34:13.013137465 +0100
@@ -474,7 +474,8 @@
virt_to_phys(page_address(cmd->SCp.buffer[1].page)+
cmd->SCp.buffer[1].offset) == endaddr; ) {
MER_PRINTK("VTOP(%p) == %08lx -> merging\n",
- cmd->SCp.buffer[1].address, endaddr);
+ page_address(cmd->SCp.buffer[1].page)+cmd->SCp.buffer[1].offset,
+ endaddr);
#if (NDEBUG & NDEBUG_MERGING)
++cnt;
#endif
@@ -987,17 +988,6 @@
#endif
/*
- * We use the host_scribble field as a pointer to the next command
- * in a queue
- */
-
- NEXT(cmd) = NULL;
- cmd->scsi_done = done;
-
- cmd->result = 0;
-
-
- /*
* Insert the cmd into the issue queue. Note that REQUEST SENSE
* commands are added to the head of the queue since any command will
* clear the contingent allegiance condition that exists and the
@@ -1018,10 +1008,31 @@
* alter queues and touch the lock.
*/
if (!IS_A_TT()) {
+ int rv;
+ /* MSch: since we get called from softirq context here, and cannot
+ * sleep safely, the return status of falcon_get_lock is now used to
+ * figure out if we could successfully lock, or need to bail out.
+ * Signal the midlevel we're unable to queue the command in this case.
+ */
oldto = atari_scsi_update_timeout(cmd, 0);
- falcon_get_lock();
+ rv = falcon_get_lock();
atari_scsi_update_timeout(cmd, oldto);
+ if (rv) {
+ local_irq_restore(flags);
+ return SCSI_MLQUEUE_HOST_BUSY;
+ }
}
+
+ /*
+ * We use the host_scribble field as a pointer to the next command
+ * in a queue
+ */
+
+ NEXT(cmd) = NULL;
+ cmd->scsi_done = done;
+
+ cmd->result = 0;
+
if (!(hostdata->issue_queue) || (cmd->cmnd[0] == REQUEST_SENSE)) {
LIST(cmd, hostdata->issue_queue);
NEXT(cmd) = hostdata->issue_queue;
@@ -1045,10 +1056,13 @@
* If we're not in an interrupt, we can call NCR5380_main()
* unconditionally, because it cannot be already running.
*/
- if (in_interrupt() || ((flags >> 8) & 7) >= 6)
+
+ /* MSch: in 2.6.19, we need to unconditionally use the task queue
+ * instead of directly starting main. Yet another side effect of
+ * the softirq business, I bet. */
+
queue_main();
- else
- NCR5380_main(NULL);
+
return 0;
}
@@ -2670,7 +2684,7 @@
* host byte of the result field to, if zero DID_ABORTED is
* used.
*
- * Returns : 0 - success, -1 on failure.
+ * Returns : SUCCESS - success, FAILED on failure.
*
* XXX - there is no way to abort the command that is currently
* connected, you have to wait for it to complete. If this is
@@ -2740,11 +2754,12 @@
local_irq_restore(flags);
cmd->scsi_done(cmd);
falcon_release_lock_if_possible( hostdata );
- return SCSI_ABORT_SUCCESS;
+ return SUCCESS;
} else {
+ /* Why is this not restoring IRQs?? */
/* local_irq_restore(flags); */
printk("scsi%d: abort of connected command failed!\n", HOSTNO);
- return SCSI_ABORT_ERROR;
+ return FAILED;
}
}
#endif
@@ -2768,7 +2783,7 @@
* yet... */
tmp->scsi_done(tmp);
falcon_release_lock_if_possible( hostdata );
- return SCSI_ABORT_SUCCESS;
+ return SUCCESS;
}
/*
@@ -2785,7 +2800,7 @@
if (hostdata->connected) {
local_irq_restore(flags);
ABRT_PRINTK("scsi%d: abort failed, command connected.\n", HOSTNO);
- return SCSI_ABORT_SNOOZE;
+ return FAILED;
}
/*
@@ -2820,7 +2835,7 @@
ABRT_PRINTK("scsi%d: aborting disconnected command.\n", HOSTNO);
if (NCR5380_select (instance, cmd, (int) cmd->tag))
- return SCSI_ABORT_BUSY;
+ return FAILED;
ABRT_PRINTK("scsi%d: nexus reestablished.\n", HOSTNO);
@@ -2847,7 +2862,7 @@
local_irq_restore(flags);
tmp->scsi_done(tmp);
falcon_release_lock_if_possible( hostdata );
- return SCSI_ABORT_SUCCESS;
+ return SUCCESS;
}
}
@@ -2871,10 +2886,9 @@
*/
falcon_release_lock_if_possible( hostdata );
- return SCSI_ABORT_NOT_RUNNING;
+ return SUCCESS;
}
-
/*
* Function : int NCR5380_reset (Scsi_Cmnd *cmd)
*
@@ -2880,7 +2894,7 @@
*
* Purpose : reset the SCSI bus.
*
- * Returns : SCSI_RESET_WAKEUP
+ * Returns : SUCCESS
*
*/
@@ -2945,15 +2959,15 @@
*/
if ((cmd = connected)) {
- ABRT_PRINTK("scsi%d: reset aborted a connected command\n", H_NO(cmd));
- cmd->result = (cmd->result & 0xffff) | (DID_RESET << 16);
+ ABRT_PRINTK("scsi%d: reset aborted a connected command, calling scsi_done() ...\n", H_NO(cmd));
+ cmd->result = (DID_RESET << 16);
cmd->scsi_done( cmd );
}
for (i = 0; (cmd = disconnected_queue); ++i) {
disconnected_queue = NEXT(cmd);
NEXT(cmd) = NULL;
- cmd->result = (cmd->result & 0xffff) | (DID_RESET << 16);
+ cmd->result = (DID_RESET << 16);
cmd->scsi_done( cmd );
}
if (i > 0)
@@ -2970,7 +2984,7 @@
* the midlevel code that the reset was SUCCESSFUL, and there is no
* need to 'wake up' the commands by a request_sense
*/
- return SCSI_RESET_SUCCESS | SCSI_RESET_BUS_RESET;
+ return SUCCESS;
#else /* 1 */
/* MSch: new-style reset handling: let the mid-level do what it can */
@@ -3018,7 +3032,8 @@
local_irq_restore(flags);
/* we did no complete reset of all commands, so a wakeup is required */
- return SCSI_RESET_WAKEUP | SCSI_RESET_BUS_RESET;
+ /* The new error handler code implicitly does this for us anyway */
+ return SUCCESS;
#endif /* 1 */
}
Signed-Off-By: schmitz@xxxxxxxxxx
Michael
-
To unsubscribe from this list: send the line "unsubscribe linux-m68k" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
