cmd64x: fix handling of address setup timings
authorBartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Mon, 18 Jan 2010 07:21:41 +0000 (07:21 +0000)
committerDavid S. Miller <davem@davemloft.net>
Tue, 19 Jan 2010 09:52:32 +0000 (01:52 -0800)
Account for the requirements of the DMA mode currently used.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/ide/cmd64x.c

index a65a691..5f80312 100644 (file)
@@ -88,9 +88,6 @@ static void cmd64x_program_timings(ide_drive_t *drive, u8 mode)
        pci_write_config_byte(dev, drwtim_regs[drive->dn],
                              (t.active << 4) | t.recover);
 
-       if (mode >= XFER_SW_DMA_0)
-               return;
-
        /*
         * The primary channel has individual address setup timing registers
         * for each drive and the hardware selects the slowest timing itself.
@@ -100,11 +97,17 @@ static void cmd64x_program_timings(ide_drive_t *drive, u8 mode)
        if (hwif->channel) {
                ide_drive_t *pair = ide_get_pair_dev(drive);
 
-               ide_set_drivedata(drive, (void *)(unsigned long)t.setup);
+               if (pair) {
+                       struct ide_timing tp;
 
-               if (pair)
-                       t.setup = max_t(u8, t.setup,
-                                       (unsigned long)ide_get_drivedata(pair));
+                       ide_timing_compute(pair, pair->pio_mode, &tp, T, 0);
+                       ide_timing_merge(&t, &tp, &t, IDE_TIMING_SETUP);
+                       if (pair->dma_mode) {
+                               ide_timing_compute(pair, pair->dma_mode,
+                                               &tp, T, 0);
+                               ide_timing_merge(&tp, &t, &t, IDE_TIMING_SETUP);
+                       }
+               }
        }
 
        if (t.setup > 5)                /* shouldn't actually happen... */