diff options
authorNithin Sujir <nsujir@broadcom.com>2013-05-21 12:57:32 +0000
committerDavid S. Miller <davem@davemloft.net>2013-05-22 14:51:11 -0700
commitfb03a43f5fb42000dcc62d91138c1c24fca609b0 (patch)
parent3ccfc1b1d2fa78f8ece83646027982916fcc794b (diff)
tg3: Ensure boot code has completed initialization before accessing hardware
After resetting the device, the driver waits for a signature to be updated to know that firmware has completed initialization. However, the call to tg3_poll_fw() is being done too late and we're writing to the GRC_MODE register before it has completely initialized, causing contention with firmware. This logic has existed since day one but is causing PCIE link to go down randomly at startup on one platform once every few hundred reboots. Move the tg3_poll_fw() up to before we write to the GRC_MODE register after reset. Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com> Signed-off-by: Michael Chan <mchan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
1 files changed, 4 insertions, 4 deletions
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index e285d7645651..158a7c90db7c 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -8911,6 +8911,10 @@ static int tg3_chip_reset(struct tg3 *tp)
tg3_halt_cpu(tp, RX_CPU_BASE);
+ err = tg3_poll_fw(tp);
+ if (err)
+ return err;
tw32(GRC_MODE, tp->grc_mode);
if (tg3_chip_rev_id(tp) == CHIPREV_ID_5705_A0) {
@@ -8941,10 +8945,6 @@ static int tg3_chip_reset(struct tg3 *tp)
tg3_ape_unlock(tp, TG3_APE_LOCK_GRC);
- err = tg3_poll_fw(tp);
- if (err)
- return err;
if (tg3_flag(tp, PCI_EXPRESS) &&