Attachment 716670 Details for Bug 928024 – [PATCH] forcedeth: Do a dma_mapping_error check after skb_frag_dma_map

[patch] [PATCH] forcedeth: Do a dma_mapping_error check after skb_frag_dma_map

-forcedeth-Do-a-dma_mapping_error-check-after-skb_frag_dma_map.patch (text/plain), 4.02 KB, created by Neil Horman on 2013-03-26 18:40:43 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Neil Horman

Created: 2013-03-26 18:40:43 UTC

Size: 4.02 KB

patch

obsolete

>From ea3ec30531c04f804659ed9883b47aaf5b247229 Mon Sep 17 00:00:00 2001
>From: Neil Horman <nhorman@tuxdriver.com>
>Date: Tue, 26 Mar 2013 14:36:56 -0400
>Subject: [PATCH] forcedeth: Do a dma_mapping_error check after
> skb_frag_dma_map
>
>This backtrace was recently reported on a 3.9 kernel:
>
>Actual results: from syslog /var/log/messsages:
>kernel: [17539.340285] ------------[ cut here ]------------
>kernel: [17539.341012] WARNING: at lib/dma-debug.c:937 check_unmap+0x493/0x960()
>kernel: [17539.341012] Hardware name: MS-7125
>kernel: [17539.341012] forcedeth 0000:00:0a.0: DMA-API: device driver failed to
>check map error[device address=0x0000000013c88000] [size=544 bytes] [mapped as
>page]
>kernel: [17539.341012] Modules linked in: fuse ebtable_nat ipt_MASQUERADE
>nf_conntrack_netbios_ns nf_conntrack_broadcast ip6table_nat nf_nat_ipv6
>ip6table_mangle ip6t_REJECT nf_conntrack_ipv6 nf_defrag_ipv6 iptable_nat
>nf_nat_ipv4 nf_nat iptable_mangle nf_conntrack_ipv4 nf_defrag_ipv4 xt_conntrack
>nf_conntrack bnep bluetooth rfkill ebtable_filter ebtables ip6table_filter
>ip6_tables snd_hda_codec_hdmi snd_cmipci snd_mpu401_uart snd_hda_intel
>snd_intel8x0 snd_opl3_lib snd_ac97_codec gameport snd_hda_codec snd_rawmidi
>ac97_bus snd_hwdep snd_seq snd_seq_device snd_pcm snd_page_alloc snd_timer snd
>k8temp soundcore serio_raw i2c_nforce2 forcedeth ata_generic pata_acpi nouveau
>video mxm_wmi wmi i2c_algo_bit drm_kms_helper ttm drm i2c_core sata_sil pata_amd
>sata_nv uinput
>kernel: [17539.341012] Pid: 17340, comm: sshd Not tainted
>3.9.0-0.rc4.git0.1.fc19.i686.PAE #1
>kernel: [17539.341012] Call Trace:
>kernel: [17539.341012]  [<c045573c>] warn_slowpath_common+0x6c/0xa0
>kernel: [17539.341012]  [<c0701953>] ? check_unmap+0x493/0x960
>kernel: [17539.341012]  [<c0701953>] ? check_unmap+0x493/0x960
>kernel: [17539.341012]  [<c04557a3>] warn_slowpath_fmt+0x33/0x40
>kernel: [17539.341012]  [<c0701953>] check_unmap+0x493/0x960
>kernel: [17539.341012]  [<c049238f>] ? sched_clock_cpu+0xdf/0x150
>kernel: [17539.341012]  [<c0701e87>] debug_dma_unmap_page+0x67/0x70
>kernel: [17539.341012]  [<f7eae8f2>] nv_unmap_txskb.isra.32+0x92/0x100
>
>Its pretty plainly the result of an skb fragment getting unmapped without having
>its initial mapping operation checked for errors.  This patch corrects that
>
>Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
>CC: "David S. Miller" <davem@davemloft.net>
>---
> drivers/net/ethernet/nvidia/forcedeth.c | 20 ++++++++++++++++++++
> 1 file changed, 20 insertions(+)
>
>diff --git a/drivers/net/ethernet/nvidia/forcedeth.c b/drivers/net/ethernet/nvidia/forcedeth.c
>index b62262c..fa9ab0a 100644
>--- a/drivers/net/ethernet/nvidia/forcedeth.c
>+++ b/drivers/net/ethernet/nvidia/forcedeth.c
>@@ -2200,6 +2200,7 @@ static netdev_tx_t nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
> 	struct ring_desc *start_tx;
> 	struct ring_desc *prev_tx;
> 	struct nv_skb_map *prev_tx_ctx;
>+	struct nv_skb_map *tmp_tx_ctx = NULL, *start_tx_ctx = NULL;
> 	unsigned long flags;
> 
> 	/* add fragments to entries count */
>@@ -2261,12 +2262,31 @@ static netdev_tx_t nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
> 		do {
> 			prev_tx = put_tx;
> 			prev_tx_ctx = np->put_tx_ctx;
>+			if (!start_tx_ctx)
>+				start_tx_ctx = tmp_tx_ctx = np->put_tx_ctx;
>+
> 			bcnt = (frag_size > NV_TX2_TSO_MAX_SIZE) ? NV_TX2_TSO_MAX_SIZE : frag_size;
> 			np->put_tx_ctx->dma = skb_frag_dma_map(
> 							&np->pci_dev->dev,
> 							frag, offset,
> 							bcnt,
> 							DMA_TO_DEVICE);
>+			if (dma_mapping_error(&np->pci_dev->dev, np->put_tx_ctx->dma)) {
>+
>+				/* Unwind the mapped fragments */
>+				do {
>+					nv_unmap_txskb(np, start_tx_ctx);
>+					if (unlikely(tmp_tx_ctx++ == np->last_tx_ctx))
>+						tmp_tx_ctx = np->first_tx_ctx;
>+				} while(tmp_tx_ctx != np->put_tx_ctx);
>+				kfree_skb(skb);
>+				np->put_tx_ctx = start_tx_ctx;
>+				u64_stats_update_begin(&np->swstats_tx_syncp);
>+				np->stat_tx_dropped++;
>+				u64_stats_update_end(&np->swstats_tx_syncp);
>+				return NETDEV_TX_OK;
>+			}
>+
> 			np->put_tx_ctx->dma_len = bcnt;
> 			np->put_tx_ctx->dma_single = 0;
> 			put_tx->buf = cpu_to_le32(np->put_tx_ctx->dma);
>-- 
>1.7.11.7
>

Actions: View | Diff

Attachments on bug 928024: 716670 | 717682