| Message ID | 20151206213926.GA30777@al |
|---|---|
| State | Awaiting Upstream, archived |
| Delegated to: | David Miller |
| Headers | show |
On 12/06/2015 03:39 PM, Peter Wu wrote: > On Sun, Dec 06, 2015 at 02:18:36PM -0600, Larry Finger wrote: >> On 12/06/2015 11:57 AM, Peter Wu wrote: >>> Free skb for received frames with a wrong checksum. >>> >>> While using the rtl8192cu driver in monitor mode, somehow 5G of memory >>> was permanently lost (observable via the Available column in `free -m`). >>> >>> Test scenario: >>> >>> ip link set down wlan1 >>> iw wlan1 set type monitor >>> ip link set up wlan1 >>> iw wlan1 set channel 11 >>> >>> Then stream a video on a smartphone on channel 11. Without this patch >>> the memory usage grows linearly with the number of received packets: >>> >>> grep MemAvailable /proc/meminfo >>> ip -s link show dev wlan1 >>> >>> Signed-off-by: Peter Wu <peter@lekensteyn.nl> >>> --- >>> Hi, >>> >>> This issue has existed since the introduction of this driver in v2.6.x, >>> using kmemleak I was about to figure out the source. There is also a >>> _rtl_usb_rx_process_agg that has similarly looking code, but that one is >>> unaffected. The pci code already frees the skb and is unaffected too. >>> >>> Tested with kernel v4.3, this patch is simply rebased on v4.4-rc3 (due >>> to changed paths). >>> >>> Kind regards, >>> Peter >>> --- >>> drivers/net/wireless/realtek/rtlwifi/usb.c | 2 ++ >>> 1 file changed, 2 insertions(+) >>> >>> diff --git a/drivers/net/wireless/realtek/rtlwifi/usb.c b/drivers/net/wireless/realtek/rtlwifi/usb.c >>> index 2721cf8..aac1ed3 100644 >>> --- a/drivers/net/wireless/realtek/rtlwifi/usb.c >>> +++ b/drivers/net/wireless/realtek/rtlwifi/usb.c >>> @@ -531,6 +531,8 @@ static void _rtl_usb_rx_process_noagg(struct ieee80211_hw *hw, >>> ieee80211_rx(hw, skb); >>> else >>> dev_kfree_skb_any(skb); >>> + } else { >>> + dev_kfree_skb_any(skb); >>> } >>> } >>> >>> >> >> Thanks for finding and fixing this memory leak. >> >> The patch is OK, but the commit message and subject are not. I do not like >> the use of the word "gigantic" in the subject. A better subject would be: >> "rtlwifi: Fix memory leak for USB device while in monitor mode". >> >> The commit message should say that a memory leak was observed, and found >> with kmemleak. If you were simply reportimg the bug, then the steps needed >> to reproduce it would be important, but as you have a fix, those steps are >> extraneous. You should also include a "Cc: Stable <stable@vger.kernel.org" >> line. When the patch is picked up for stable kernels, it will be necessary >> to rebase the patch to compensate for the directory change. > > I have done some additional testing using QEMU and USB passthrough and > can report that the leak is not limited to monitor mode. The commit > message is adjusted for that. This issue is pretty bad, I previously hit > 5GB within an hour with low traffic, this VM with 256M RAM panics after > 5 minutes in managed mode. (Remote denial of service, heh.) > > Originally I had the Cc: stable line added, but the SubmittingPatches > document seems to discourage that for networking. Added it again. > > Here is the updated patch, hopefully addressing your concerns. Feel free > to modify it as you find appropriate. > > Peter > -- >>From a2c5fec1789bef48671d643ea7ecd0244d1e0246 Mon Sep 17 00:00:00 2001 > From: Peter Wu <peter@lekensteyn.nl> > Date: Sun, 6 Dec 2015 17:59:41 +0100 > Subject: [PATCH] rtlwifi: fix memory leak for USB device > > Free skb for received frames with a wrong checksum. This can happen > pretty rapidly, exhausting all memory. > > This fixes a memleak (detected with kmemleak). Originally found while > using monitor mode, but it also appears during managed mode (once the > link is up). > > Cc: stable@vger.kernel.org > Signed-off-by: Peter Wu <peter@lekensteyn.nl> > --- This version is better, but you submitted it incorrectly. Sen it as [PATCH V2] and cut out all extraneous messaging. Larry -- 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
Peter Wu <peter@lekensteyn.nl> writes: > Originally I had the Cc: stable line added, but the SubmittingPatches > document seems to discourage that for networking. Added it again. Yeah, stable wireless patches are handled differently from rest of the networking subsystem. It would be great if somebody could update the documentation.
On 12/06/2015 09:39 PM, Peter Wu wrote: >>> While using the rtl8192cu driver in monitor mode, somehow 5G of memory >>> was permanently lost (observable via the Available column in `free -m`). >>> >>> This issue has existed since the introduction of this driver in v2.6.x, One more reason to switch to rtl8xxxu as soon as possible... bruno -- 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/wireless/realtek/rtlwifi/usb.c b/drivers/net/wireless/realtek/rtlwifi/usb.c index 2721cf8..aac1ed3 100644 --- a/drivers/net/wireless/realtek/rtlwifi/usb.c +++ b/drivers/net/wireless/realtek/rtlwifi/usb.c @@ -531,6 +531,8 @@ static void _rtl_usb_rx_process_noagg(struct ieee80211_hw *hw, ieee80211_rx(hw, skb); else dev_kfree_skb_any(skb); + } else { + dev_kfree_skb_any(skb); } }