| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <51D29391.4080000@broadcom.com>
Date: Tue, 2 Jul 2013 10:47:13 +0200
From: "Arend van Spriel" <arend@...adcom.com>
To: "Larry Finger" <Larry.Finger@...inger.net>
cc: linville@...driver.com, linux-wireless@...r.kernel.org,
netdev@...r.kernel.org, Stable <stable@...r.kernel.org>,
"Brett Rudley" <brudley@...adcom.com>,
"Franky (Zhenhui) Lin" <frankyl@...adcom.com>,
"Hante Meuleman" <meuleman@...adcom.com>,
brcm80211-dev-list@...adcom.com,
"Neil Horman" <nhorman@...driver.com>
Subject: Re: [PATCH] brcmsmac: Fix WARNING caused by lack of calls to
dma_mapping_error()
On 07/01/2013 11:08 PM, Larry Finger wrote:
> The driver fails to check the results of DMA mapping in twp places, which results
> in the following warning:
>
> [ 28.078515] ------------[ cut here ]------------
> [ 28.078529] WARNING: at lib/dma-debug.c:937 check_unmap+0x47e/0x930()
> [ 28.078533] bcma-pci-bridge 0000:0e:00.0: DMA-API: device driver failed to check map error[device address=0x00000000b5d60d6c] [size=1876 bytes] [mapped as
> single]
> [ 28.078536] Modules linked in: bnep bluetooth vboxpci(O) vboxnetadp(O) vboxnetflt(O) vboxdrv(O) ipv6 b43 brcmsmac rtl8192cu rtl8192c_common rtlwifi mac802
> 11 brcmutil cfg80211 snd_hda_codec_conexant rng_core snd_hda_intel kvm_amd snd_hda_codec ssb kvm mmc_core snd_pcm snd_seq snd_timer snd_seq_device snd k8temp
> cordic joydev serio_raw hwmon sr_mod sg pcmcia pcmcia_core soundcore cdrom i2c_nforce2 i2c_core forcedeth bcma snd_page_alloc autofs4 ext4 jbd2 mbcache crc1
> 6 scsi_dh_alua scsi_dh_hp_sw scsi_dh_rdac scsi_dh_emc scsi_dh ata_generic pata_amd
> [ 28.078602] CPU: 1 PID: 2570 Comm: NetworkManager Tainted: G O 3.10.0-rc7-wl+ #42
> [ 28.078605] Hardware name: Hewlett-Packard HP Pavilion dv2700 Notebook PC/30D6, BIOS F.27 11/27/2008
> [ 28.078607] 0000000000000009 ffff8800bbb03ad8 ffffffff8144f898 ffff8800bbb03b18
> [ 28.078612] ffffffff8103e1eb 0000000000000002 ffff8800b719f480 ffff8800b7b9c010
> [ 28.078617] ffffffff824204c0 ffffffff81754d57 0000000000000754 ffff8800bbb03b78
> [ 28.078622] Call Trace:
> [ 28.078624] <IRQ> [<ffffffff8144f898>] dump_stack+0x19/0x1b
> [ 28.078634] [<ffffffff8103e1eb>] warn_slowpath_common+0x6b/0xa0
> [ 28.078638] [<ffffffff8103e2c1>] warn_slowpath_fmt+0x41/0x50
> [ 28.078650] [<ffffffff8122d7ae>] check_unmap+0x47e/0x930
> [ 28.078655] [<ffffffff8122de4c>] debug_dma_unmap_page+0x5c/0x70
> [ 28.078679] [<ffffffffa04a808c>] dma64_getnextrxp+0x10c/0x190 [brcmsmac]
> [ 28.078691] [<ffffffffa04a9042>] dma_rx+0x62/0x240 [brcmsmac]
> [ 28.078707] [<ffffffffa0479101>] brcms_c_dpc+0x211/0x9d0 [brcmsmac]
> [ 28.078717] [<ffffffffa046d927>] ? brcms_dpc+0x27/0xf0 [brcmsmac]
> [ 28.078731] [<ffffffffa046d947>] brcms_dpc+0x47/0xf0 [brcmsmac]
> [ 28.078736] [<ffffffff81047dcc>] tasklet_action+0x6c/0xf0
> --snip--
> [ 28.078974] [<ffffffff813891bd>] SyS_sendmsg+0xd/0x20
> [ 28.078979] [<ffffffff81455c24>] tracesys+0xdd/0xe2
> [ 28.078982] ---[ end trace 6164d1a08148e9c8 ]---
> [ 28.078984] Mapped at:
> [ 28.078985] [<ffffffff8122c8fd>] debug_dma_map_page+0x9d/0x150
> [ 28.078989] [<ffffffffa04a9322>] dma_rxfill+0x102/0x3d0 [brcmsmac]
> [ 28.079001] [<ffffffffa047a13d>] brcms_c_init+0x87d/0x1100 [brcmsmac]
> [ 28.079010] [<ffffffffa046d851>] brcms_init+0x21/0x30 [brcmsmac]
> [ 28.079018] [<ffffffffa04786e0>] brcms_c_up+0x150/0x430 [brcmsmac]
>
> As the patch adds a new failure mechanism to dma_rxfill(). When I changed the
> comment at the start of the routine to add that information, I also polished
> the wording.
Thanks, Larry
I think there a bugzilla issue for this that you might want to reference:
https://bugzilla.kernel.org/show_bug.cgi?id=60261
> Signed-off-by: Larry Finger <Larry.Finger@...inger.net>
> Cc: Stable <stable@...r.kernel.org>
> Cc: Brett Rudley <brudley@...adcom.com>
> Cc: Arend van Spriel <arend@...adcom.com>
> Cc: Franky (Zhenhui) Lin <frankyl@...adcom.com>
> Cc: Hante Meuleman <meuleman@...adcom.com>
> Cc: brcm80211-dev-list@...adcom.com
> ---
>
> drivers/net/wireless/brcm80211/brcmsmac/dma.c | 12 ++++++++----
> 1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/wireless/brcm80211/brcmsmac/dma.c b/drivers/net/wireless/brcm80211/brcmsmac/dma.c
> index 1860c57..6260571 100644
> --- a/drivers/net/wireless/brcm80211/brcmsmac/dma.c
> +++ b/drivers/net/wireless/brcm80211/brcmsmac/dma.c
> @@ -1015,9 +1015,10 @@ static bool dma64_txidle(struct dma_info *di)
>
> /*
> * post receive buffers
> - * return false is refill failed completely and ring is empty this will stall
> - * the rx dma and user might want to call rxfill again asap. This unlikely
> - * happens on memory-rich NIC, but often on memory-constrained dongle
> + * return false is refill failed completely or dma mapping failed. The ring
Missed some polishing opportunity: return false is refill... -> return
false if refill...
> + * is empty, which will stall the rx dma and user might want to call rxfill
> + * again asap. This is unlikely to happen on a memory-rich NIC, but often on
> + * memory-constrained dongle
> */
> bool dma_rxfill(struct dma_pub *pub)
> {
> @@ -1078,6 +1079,8 @@ bool dma_rxfill(struct dma_pub *pub)
>
> pa = dma_map_single(di->dmadev, p->data, di->rxbufsize,
> DMA_FROM_DEVICE);
> + if (dma_mapping_error(di->dmadev,pa))
> + return false;
>
> /* save the free packet pointer */
> di->rxp[rxout] = p;
> @@ -1284,7 +1287,8 @@ static void dma_txenq(struct dma_info *di, struct sk_buff *p)
>
> /* get physical address of buffer start */
> pa = dma_map_single(di->dmadev, data, len, DMA_TO_DEVICE);
> -
> + if(dma_mapping_error(di->dmadev, pa))
> + return;
I think the caller(s) should be informed or silently free the sk_buff
here so we are not leaking it.
> /* With a DMA segment list, Descriptor table is filled
> * using the segment list instead of looping over
> * buffers in multi-chain DMA. Therefore, EOF for SGLIST
>
Regards,
Arend
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists