| Message ID | 1447671527-28905-1-git-send-email-rajesh.borundia@qlogic.com |
|---|---|
| State | Rejected, archived |
| Delegated to: | David Miller |
| Headers | show |
From: Rajesh Borundia <rajesh.borundia@qlogic.com> Date: Mon, 16 Nov 2015 05:58:47 -0500 > + } else { > + if (atomic_read(&mbx->rsp_status) != rsp_status) > + qlcnic_83xx_notify_mbx_response(mbx); Here we go with more of some "We don't understand what atomic_t's actually do"... atomic_read() does _nothing_ It doesn't enforce ordering of the read of the variable wrt. other memory accesses or anything like that. It doesn't make the read "atomic". All it does is convert an "atomic_t" into an "int", and such a conversion is necessary if the implementation makes use of some of the atomic_t value for implementing atomicity (such as putting a spinlock in the upper byte of the value, etc.) Therefore, if you are using an atomic_t to order stores and reads of an integer value. That's not what they do, and your usage is wrong. If you are assuming that atomic_t's store a full 32-bit signed integer, all 32-bits of which which are unmolested by the implementation, then your usage is wrong. The implementation may use any number of bits of the atomic_t value for it's own usage. As a result, these uses in the qlogic driver are completely wrong and give a false impression as to what the atomic_read() function does and what atomic_t variables are to be used for. Please correct this, and then respin your patch on top of that. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/net/ethernet/qlogic/qlcnic/qlcnic.h b/drivers/net/ethernet/qlogic/qlcnic/qlcnic.h index d6696cf..371c266 100644 --- a/drivers/net/ethernet/qlogic/qlcnic/qlcnic.h +++ b/drivers/net/ethernet/qlogic/qlcnic/qlcnic.h @@ -566,6 +566,7 @@ struct qlcnic_adapter_stats { u64 tx_dma_map_error; u64 spurious_intr; u64 mac_filter_limit_overrun; + u64 mbx_spurious_intr; }; /* diff --git a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_83xx_hw.c b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_83xx_hw.c index 9f0bdd9..865f21b 100644 --- a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_83xx_hw.c +++ b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_83xx_hw.c @@ -2338,9 +2338,9 @@ static void qlcnic_83xx_handle_link_aen(struct qlcnic_adapter *adapter, static irqreturn_t qlcnic_83xx_handle_aen(int irq, void *data) { + u32 mask, resp, event, rsp_status = QLC_83XX_MBX_RESPONSE_ARRIVED; struct qlcnic_adapter *adapter = data; struct qlcnic_mailbox *mbx; - u32 mask, resp, event; unsigned long flags; mbx = adapter->ahw->mailbox; @@ -2350,10 +2350,14 @@ static irqreturn_t qlcnic_83xx_handle_aen(int irq, void *data) goto out; event = readl(QLCNIC_MBX_FW(adapter->ahw, 0)); - if (event & QLCNIC_MBX_ASYNC_EVENT) + if (event & QLCNIC_MBX_ASYNC_EVENT) { __qlcnic_83xx_process_aen(adapter); - else - qlcnic_83xx_notify_mbx_response(mbx); + } else { + if (atomic_read(&mbx->rsp_status) != rsp_status) + qlcnic_83xx_notify_mbx_response(mbx); + else + adapter->stats.mbx_spurious_intr++; + } out: mask = QLCRDX(adapter->ahw, QLCNIC_DEF_INT_MASK); diff --git a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_ethtool.c b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_ethtool.c index 494e810..0a2318c 100644 --- a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_ethtool.c +++ b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_ethtool.c @@ -59,7 +59,8 @@ static const struct qlcnic_stats qlcnic_gstrings_stats[] = { QLC_OFF(stats.mac_filter_limit_overrun)}, {"spurious intr", QLC_SIZEOF(stats.spurious_intr), QLC_OFF(stats.spurious_intr)}, - + {"mbx spurious intr", QLC_SIZEOF(stats.mbx_spurious_intr), + QLC_OFF(stats.mbx_spurious_intr)}, }; static const char qlcnic_device_gstrings_stats[][ETH_GSTRING_LEN] = {
o While the driver is in the middle of a MB completion processing and it receives a spurious MB interrupt, it is mistaken as a good MB completion interrupt leading to premature completion of the next MB request. Fix the driver to guard against this by checking the current state of MB processing and ignore the spurious interrupt. Also added a stats counter to record this condition. Signed-off-by: Rajesh Borundia <rajesh.borundia@qlogic.com> --- drivers/net/ethernet/qlogic/qlcnic/qlcnic.h | 1 + .../net/ethernet/qlogic/qlcnic/qlcnic_83xx_hw.c | 12 ++++++++---- .../net/ethernet/qlogic/qlcnic/qlcnic_ethtool.c | 3 ++- 3 files changed, 11 insertions(+), 5 deletions(-)