From 1319ebadf185933e6b7ff95211d3cef9004e9754 Mon Sep 17 00:00:00 2001 From: Lennert Buytenhek Date: Wed, 29 Apr 2009 11:57:34 +0000 Subject: mv643xx_eth: OOM handling fixes Currently, when OOM occurs during rx ring refill, mv643xx_eth will get into an infinite loop, due to the refill function setting the OOM bit but not clearing the 'rx refill needed' bit for this queue, while the calling function (the NAPI poll handler) will call the refill function in a loop until the 'rx refill needed' bit goes off, without checking the OOM bit. This patch fixes this by checking the OOM bit in the NAPI poll handler before attempting to do rx refill. This means that once OOM occurs, we won't try to do any memory allocations again until the next invocation of the poll handler. While we're at it, change the OOM flag to be a single bit instead of one bit per receive queue since OOM is a system state rather than a per-queue state, and cancel the OOM timer on entry to the NAPI poll handler if it's running to prevent it from firing when we've already come out of OOM. Signed-off-by: Lennert Buytenhek Cc: stable@kernel.org Signed-off-by: David S. Miller --- drivers/net/mv643xx_eth.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-) (limited to 'drivers/net/mv643xx_eth.c') diff --git a/drivers/net/mv643xx_eth.c b/drivers/net/mv643xx_eth.c index b3185bf2c15..038beff7da8 100644 --- a/drivers/net/mv643xx_eth.c +++ b/drivers/net/mv643xx_eth.c @@ -393,12 +393,12 @@ struct mv643xx_eth_private { struct work_struct tx_timeout_task; struct napi_struct napi; + u8 oom; u8 work_link; u8 work_tx; u8 work_tx_end; u8 work_rx; u8 work_rx_refill; - u8 work_rx_oom; int skb_size; struct sk_buff_head rx_recycle; @@ -661,7 +661,7 @@ static int rxq_refill(struct rx_queue *rxq, int budget) dma_get_cache_alignment() - 1); if (skb == NULL) { - mp->work_rx_oom |= 1 << rxq->index; + mp->oom = 1; goto oom; } @@ -2167,8 +2167,10 @@ static int mv643xx_eth_poll(struct napi_struct *napi, int budget) mp = container_of(napi, struct mv643xx_eth_private, napi); - mp->work_rx_refill |= mp->work_rx_oom; - mp->work_rx_oom = 0; + if (unlikely(mp->oom)) { + mp->oom = 0; + del_timer(&mp->rx_oom); + } work_done = 0; while (work_done < budget) { @@ -2182,8 +2184,10 @@ static int mv643xx_eth_poll(struct napi_struct *napi, int budget) continue; } - queue_mask = mp->work_tx | mp->work_tx_end | - mp->work_rx | mp->work_rx_refill; + queue_mask = mp->work_tx | mp->work_tx_end | mp->work_rx; + if (likely(!mp->oom)) + queue_mask |= mp->work_rx_refill; + if (!queue_mask) { if (mv643xx_eth_collect_events(mp)) continue; @@ -2204,7 +2208,7 @@ static int mv643xx_eth_poll(struct napi_struct *napi, int budget) txq_maybe_wake(mp->txq + queue); } else if (mp->work_rx & queue_mask) { work_done += rxq_process(mp->rxq + queue, work_tbd); - } else if (mp->work_rx_refill & queue_mask) { + } else if (!mp->oom && (mp->work_rx_refill & queue_mask)) { work_done += rxq_refill(mp->rxq + queue, work_tbd); } else { BUG(); @@ -2212,7 +2216,7 @@ static int mv643xx_eth_poll(struct napi_struct *napi, int budget) } if (work_done < budget) { - if (mp->work_rx_oom) + if (mp->oom) mod_timer(&mp->rx_oom, jiffies + (HZ / 10)); napi_complete(napi); wrlp(mp, INT_MASK, INT_TX_END | INT_RX | INT_EXT); @@ -2372,7 +2376,7 @@ static int mv643xx_eth_open(struct net_device *dev) rxq_refill(mp->rxq + i, INT_MAX); } - if (mp->work_rx_oom) { + if (mp->oom) { mp->rx_oom.expires = jiffies + (HZ / 10); add_timer(&mp->rx_oom); } -- cgit v1.2.3 From 93af7aca44f0e82e67bda10a0fb73d383edcc8bd Mon Sep 17 00:00:00 2001 From: Lennert Buytenhek Date: Wed, 29 Apr 2009 11:58:18 +0000 Subject: mv643xx_eth: 64bit mib counter read fix On several mv643xx_eth hardware versions, the two 64bit mib counters for 'good octets received' and 'good octets sent' are actually 32bit counters, and reading from the upper half of the register has the same effect as reading from the lower half of the register: an atomic read-and-clear of the entire 32bit counter value. This can under heavy traffic occasionally lead to small numbers being added to the upper half of the 64bit mib counter even though no 32bit wrap has occured. Since we poll the mib counters at least every 30 seconds anyway, we might as well just skip the reads of the upper halves of the hardware counters without breaking the stats, which this patch does. Signed-off-by: Lennert Buytenhek Cc: stable@kernel.org Signed-off-by: David S. Miller --- drivers/net/mv643xx_eth.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers/net/mv643xx_eth.c') diff --git a/drivers/net/mv643xx_eth.c b/drivers/net/mv643xx_eth.c index 038beff7da8..a400d7115f7 100644 --- a/drivers/net/mv643xx_eth.c +++ b/drivers/net/mv643xx_eth.c @@ -1255,7 +1255,6 @@ static void mib_counters_update(struct mv643xx_eth_private *mp) spin_lock_bh(&mp->mib_counters_lock); p->good_octets_received += mib_read(mp, 0x00); - p->good_octets_received += (u64)mib_read(mp, 0x04) << 32; p->bad_octets_received += mib_read(mp, 0x08); p->internal_mac_transmit_err += mib_read(mp, 0x0c); p->good_frames_received += mib_read(mp, 0x10); @@ -1269,7 +1268,6 @@ static void mib_counters_update(struct mv643xx_eth_private *mp) p->frames_512_to_1023_octets += mib_read(mp, 0x30); p->frames_1024_to_max_octets += mib_read(mp, 0x34); p->good_octets_sent += mib_read(mp, 0x38); - p->good_octets_sent += (u64)mib_read(mp, 0x3c) << 32; p->good_frames_sent += mib_read(mp, 0x40); p->excessive_collision += mib_read(mp, 0x44); p->multicast_frames_sent += mib_read(mp, 0x48); -- cgit v1.2.3