Hi Larry,
I'm experiencing an issue with rtl8192se since 3.1-ish timeframe where
the machine becomes completely unresponsive and only a reboot helps
the situation. I think the issue has to do with the rtl8192se wireless
driver because if I connect the machine through ethernet, it runs pretty
smoothly.
Today, I left the machine on the vt console in the expectation of an
oops or something to appear in the logs and logged into it from another
machine over ssh.
After a while, the unresponsiveness happened and the box didn't
react to keyboard input except sysrq with which I was able to do the
show-backtrace-all-active-cpus(L) thing and attached is a partial
screen cap of that. It looks like the stuck-up happens somewhere in
rtl_lps_leave() along the rtl92s_phy_set_rf_power_state() path but the
register dump is missing with the exact %rIP.
Anyway, pls take a look and let me know if it rings any bells. I'll
continue trying to debug the issue, maybe I should bisect it if nothing
else pops up.
TIA.
--
Regards/Gruss,
Boris.
On Thu, Sep 15, 2011 at 10:23:29AM -0500, Larry Finger wrote:
> Would you also try loading rtl8192se with the "ips=0" option? As power
> save is implicated in your traceback, that may help. I will be trying
> "swlps=0 ips=0".
Ok, "ips=0" seems to fix the issue ... almost. I say, almost because I
had only one hang so far for running the box for a day today. Will try
together with "swlps=0" next week.
Thanks.
--
Regards/Gruss,
Boris.
On 09/15/2011 07:44 AM, Borislav Petkov wrote:
> Hi Larry,
>
> I'm experiencing an issue with rtl8192se since 3.1-ish timeframe where
> the machine becomes completely unresponsive and only a reboot helps
> the situation. I think the issue has to do with the rtl8192se wireless
> driver because if I connect the machine through ethernet, it runs pretty
> smoothly.
>
> Today, I left the machine on the vt console in the expectation of an
> oops or something to appear in the logs and logged into it from another
> machine over ssh.
>
> After a while, the unresponsiveness happened and the box didn't
> react to keyboard input except sysrq with which I was able to do the
> show-backtrace-all-active-cpus(L) thing and attached is a partial
> screen cap of that. It looks like the stuck-up happens somewhere in
> rtl_lps_leave() along the rtl92s_phy_set_rf_power_state() path but the
> register dump is missing with the exact %rIP.
>
> Anyway, pls take a look and let me know if it rings any bells. I'll
> continue trying to debug the issue, maybe I should bisect it if nothing
> else pops up.
Borislav,
Thanks for the report. I have been running rtl8192se for the past few days and I
have also noticed two such system freezes, but not been able to capture any
info. As I have recently made many changes in my system recently, I did not know
what might be the cause, but rtl8192se is certainly on the suspect list.
I have added Chaoming Li to the Cc list. I will send him the screen photo
separately.
Some questions:
I expect that you are running a mainline kernel from Linus's tree. If not,
please let me know. Mine is 3.1-rc4 from the wireless-testing tree. I don't
recall any changes in out "next-flavored" version that are not in 3.1.
Which flavor of RTL8192SE card do you have? The one I'm running shows as
"Realtek Semiconductor Co., Ltd. RTL8191SEvB Wireless LAN Controller [10ec:8172]
(rev 10)", but I have two others. The differences are in the number of TX and RX
streams. Mine is the 1x2 variety.
How frequently do your freezes occur? As I said before, I have only had two in 2
or 3 days, which would make bisection tricky.
I see from the dump that you have x86_64 architecture. How many CPUs and how
fast? There is one questionable report of problems on a box with an 8-way fast
processor. That was on initialization and is not the same, but may indicate a
problem. My system has a dual AMD CPU at 2.0 GHz.
Would you also try loading rtl8192se with the "ips=0" option? As power save is
implicated in your traceback, that may help. I will be trying "swlps=0 ips=0".
Thanks for the report,
Larry
On Thu, Sep 15, 2011 at 10:23:29AM -0500, Larry Finger wrote:
> Thanks for the report. I have been running rtl8192se for the past
> few days and I have also noticed two such system freezes, but not
Ah ok, so it's not only me seeing this.
> I expect that you are running a mainline kernel from Linus's tree.
It is Linus' tree: v3.1-rc6-10-g003f6c9
> If not, please let me know. Mine is 3.1-rc4 from the
> wireless-testing tree. I don't recall any changes in out
> "next-flavored" version that are not in 3.1.
>
> Which flavor of RTL8192SE card do you have? The one I'm running
> shows as "Realtek Semiconductor Co., Ltd. RTL8191SEvB Wireless LAN
> Controller [10ec:8172] (rev 10)", but I have two others. The
> differences are in the number of TX and RX streams. Mine is the 1x2
> variety.
lspci says:
03:00.0 Network controller [0280]: Realtek Semiconductor Co., Ltd. RTL8191SEvB Wireless LAN Controller [10ec:8172] (rev 10)
so it is exactly the same as yours.
> How frequently do your freezes occur? As I said before, I have only
> had two in 2 or 3 days, which would make bisection tricky.
Well, I can reproduce it pretty reliably: it happens shortly after I
up the iface and establish the WIFI connection with wpa_supplicant. A
couple of minutes after that, more or less, the box grinds down to a
halt.
> I see from the dump that you have x86_64 architecture. How many CPUs
> and how fast? There is one questionable report of problems on a box
> with an 8-way fast processor. That was on initialization and is not
> the same, but may indicate a problem. My system has a dual AMD CPU
> at 2.0 GHz.
I don't think that has any effect on the wifi iface but here it is: dual
core K8 laptop:
...
processor : 1
vendor_id : AuthenticAMD
cpu family : 15
model : 107
model name : AMD Turion(tm) Neo X2 Dual Core Processor L625
stepping : 2
cpu MHz : 800.000
this is of cource the lowest P-state freq - P0 is 1.6GHz.
> Would you also try loading rtl8192se with the "ips=0" option? As
> power save is implicated in your traceback, that may help. I will be
> trying "swlps=0 ips=0".
Ok, I'll run both just in case and let you know.
Thanks for looking into this.
--
Regards/Gruss,
Boris.