aspeed/watchdog: Correct timeout value

The driver was using milliseconds and programming it into a register
which takes ticks of the watchdog clock, which runs at 1MHz. This meant
we were off by 1000 with the desired value.

When 06985289d452 ("watchdog: Implement generic watchdog_reset()
version") was added the aspeed board would leave the watchdog running,
causing it to bite before u-boot was done.

Discovered by booting in qemu:

  $ qemu-system-arm -M ast2500-evb -drive file=test.img,format=raw,if=mtd -nographic -no-reboot -d cpu_reset

  U-Boot 2019.07-rc3-00091-g2253e40caef5 (Jun 06 2019 - 16:53:23 +0930)

  Model: Aspeed BMC
  DRAM:  496 MiB
  WDT:   Started with servicing (60s timeout)
  MMC:
  In:    serial@1e784000
  Out:   serial@1e784000
  Err:   serial@1e784000
  Watchdog timer expired.

Fixes: 06985289d452 ("watchdog: Implement generic watchdog_reset() version")
Signed-off-by: Joel Stanley <joel@jms.id.au>

diff --git a/drivers/watchdog/ast_wdt.c b/drivers/watchdog/ast_wdt.c
index 523484b1fff945c335d2da11a8b977f93d7e6633..d344d54aee828fdeb2ff72dbe79ec557fc998d52 100644 (file)
--- a/drivers/watchdog/ast_wdt.c
+++ b/drivers/watchdog/ast_wdt.c
@@ -23,6 +23,12 @@ static int ast_wdt_start(struct udevice *dev, u64 timeout, ulong flags)
        ulong driver_data = dev_get_driver_data(dev);
        u32 reset_mode = ast_reset_mode_from_flags(flags);
 
+       /* 32 bits at 1MHz is 4294967ms */
+       timeout = min_t(u64, timeout, 4294967);
+
+       /* WDT counts in ticks of 1MHz clock. 1ms / 1e3 * 1e6 */
+       timeout *= 1000;
+
        clrsetbits_le32(&priv->regs->ctrl,
                        WDT_CTRL_RESET_MASK << WDT_CTRL_RESET_MODE_SHIFT,
                        reset_mode << WDT_CTRL_RESET_MODE_SHIFT);