From: Johannes Berg Date: Tue, 18 Nov 2008 00:47:21 +0000 (+0100) Subject: iwlagn: fix RX skb alignment X-Git-Tag: v2.6.28-rc6~8^2~4^2 X-Git-Url: http://pilppa.com/gitweb/?a=commitdiff_plain;h=4018517a1a69a85c3d61b20fa02f187b80773137;p=linux-2.6-omap-h63xx.git iwlagn: fix RX skb alignment So I dug deeper into the DMA problems I had with iwlagn and a kind soul helped me in that he said something about pci-e alignment and mentioned the iwl_rx_allocate function to check for crossing 4KB boundaries. Since there's 8KB A-MPDU support, crossing 4k boundaries didn't seem like something the device would fail with, but when I looked into the function for a minute anyway I stumbled over this little gem: BUG_ON(rxb->dma_addr & (~DMA_BIT_MASK(36) & 0xff)); Clearly, that is a totally bogus check, one would hope the compiler removes it entirely. (Think about it) After fixing it, I obviously ran into it, nothing guarantees the alignment the way you want it, because of the way skbs and their headroom are allocated. I won't explain that here nor double-check that I'm right, that goes beyond what most of the CC'ed people care about. So then I came up with the patch below, and so far my system has survived minutes with 64K pages, when it would previously fail in seconds. And I haven't seen a single instance of the TX bug either. But when you see the patch it'll be pretty obvious to you why. This should fix the following reported kernel bugs: http://bugzilla.kernel.org/show_bug.cgi?id=11596 http://bugzilla.kernel.org/show_bug.cgi?id=11393 http://bugzilla.kernel.org/show_bug.cgi?id=11983 I haven't checked if there are any elsewhere, but I suppose RHBZ will have a few instances too... I'd like to ask anyone who is CC'ed (those are people I know ran into the bug) to try this patch. I am convinced that this patch is correct in spirit, but I haven't understood why, for example, there are so many unmap calls. I'm not entirely convinced that this is the only bug leading to the TX reply errors. Signed-off-by: Johannes Berg Signed-off-by: John W. Linville --- diff --git a/drivers/net/wireless/iwlwifi/iwl-agn.c b/drivers/net/wireless/iwlwifi/iwl-agn.c index 6751bb2b8ae..444c5cc05f0 100644 --- a/drivers/net/wireless/iwlwifi/iwl-agn.c +++ b/drivers/net/wireless/iwlwifi/iwl-agn.c @@ -1384,7 +1384,7 @@ void iwl_rx_handle(struct iwl_priv *priv) rxq->queue[i] = NULL; - pci_dma_sync_single_for_cpu(priv->pci_dev, rxb->dma_addr, + pci_dma_sync_single_for_cpu(priv->pci_dev, rxb->aligned_dma_addr, priv->hw_params.rx_buf_size, PCI_DMA_FROMDEVICE); pkt = (struct iwl_rx_packet *)rxb->skb->data; @@ -1436,8 +1436,8 @@ void iwl_rx_handle(struct iwl_priv *priv) rxb->skb = NULL; } - pci_unmap_single(priv->pci_dev, rxb->dma_addr, - priv->hw_params.rx_buf_size, + pci_unmap_single(priv->pci_dev, rxb->real_dma_addr, + priv->hw_params.rx_buf_size + 256, PCI_DMA_FROMDEVICE); spin_lock_irqsave(&rxq->lock, flags); list_add_tail(&rxb->list, &priv->rxq.rx_used); diff --git a/drivers/net/wireless/iwlwifi/iwl-dev.h b/drivers/net/wireless/iwlwifi/iwl-dev.h index c018121085e..9966d4e384c 100644 --- a/drivers/net/wireless/iwlwifi/iwl-dev.h +++ b/drivers/net/wireless/iwlwifi/iwl-dev.h @@ -89,7 +89,8 @@ extern struct iwl_cfg iwl5100_abg_cfg; #define DEFAULT_LONG_RETRY_LIMIT 4U struct iwl_rx_mem_buffer { - dma_addr_t dma_addr; + dma_addr_t real_dma_addr; + dma_addr_t aligned_dma_addr; struct sk_buff *skb; struct list_head list; }; diff --git a/drivers/net/wireless/iwlwifi/iwl-rx.c b/drivers/net/wireless/iwlwifi/iwl-rx.c index 7cde9d76ff5..0509c16dbe7 100644 --- a/drivers/net/wireless/iwlwifi/iwl-rx.c +++ b/drivers/net/wireless/iwlwifi/iwl-rx.c @@ -204,7 +204,7 @@ int iwl_rx_queue_restock(struct iwl_priv *priv) list_del(element); /* Point to Rx buffer via next RBD in circular buffer */ - rxq->bd[rxq->write] = iwl_dma_addr2rbd_ptr(priv, rxb->dma_addr); + rxq->bd[rxq->write] = iwl_dma_addr2rbd_ptr(priv, rxb->aligned_dma_addr); rxq->queue[rxq->write] = rxb; rxq->write = (rxq->write + 1) & RX_QUEUE_MASK; rxq->free_count--; @@ -251,7 +251,7 @@ void iwl_rx_allocate(struct iwl_priv *priv) rxb = list_entry(element, struct iwl_rx_mem_buffer, list); /* Alloc a new receive buffer */ - rxb->skb = alloc_skb(priv->hw_params.rx_buf_size, + rxb->skb = alloc_skb(priv->hw_params.rx_buf_size + 256, __GFP_NOWARN | GFP_ATOMIC); if (!rxb->skb) { if (net_ratelimit()) @@ -266,9 +266,17 @@ void iwl_rx_allocate(struct iwl_priv *priv) list_del(element); /* Get physical address of RB/SKB */ - rxb->dma_addr = - pci_map_single(priv->pci_dev, rxb->skb->data, - priv->hw_params.rx_buf_size, PCI_DMA_FROMDEVICE); + rxb->real_dma_addr = pci_map_single( + priv->pci_dev, + rxb->skb->data, + priv->hw_params.rx_buf_size + 256, + PCI_DMA_FROMDEVICE); + /* dma address must be no more than 36 bits */ + BUG_ON(rxb->real_dma_addr & ~DMA_BIT_MASK(36)); + /* and also 256 byte aligned! */ + rxb->aligned_dma_addr = ALIGN(rxb->real_dma_addr, 256); + skb_reserve(rxb->skb, rxb->aligned_dma_addr - rxb->real_dma_addr); + list_add_tail(&rxb->list, &rxq->rx_free); rxq->free_count++; } @@ -300,8 +308,8 @@ void iwl_rx_queue_free(struct iwl_priv *priv, struct iwl_rx_queue *rxq) for (i = 0; i < RX_QUEUE_SIZE + RX_FREE_BUFFERS; i++) { if (rxq->pool[i].skb != NULL) { pci_unmap_single(priv->pci_dev, - rxq->pool[i].dma_addr, - priv->hw_params.rx_buf_size, + rxq->pool[i].real_dma_addr, + priv->hw_params.rx_buf_size + 256, PCI_DMA_FROMDEVICE); dev_kfree_skb(rxq->pool[i].skb); } @@ -354,8 +362,8 @@ void iwl_rx_queue_reset(struct iwl_priv *priv, struct iwl_rx_queue *rxq) * to an SKB, so we need to unmap and free potential storage */ if (rxq->pool[i].skb != NULL) { pci_unmap_single(priv->pci_dev, - rxq->pool[i].dma_addr, - priv->hw_params.rx_buf_size, + rxq->pool[i].real_dma_addr, + priv->hw_params.rx_buf_size + 256, PCI_DMA_FROMDEVICE); priv->alloc_rxb_skb--; dev_kfree_skb(rxq->pool[i].skb);