mjhjmtu

I have a question regarding PAM and the possibility that it throttles too many 'auth' attempts if they happen in a short time.

I have posted my question on the PAM mailing list, but to no avail. I figured I'd share it here too, in case you guys have some insight.

Long story short: I am not seeking to increase the number of attempts after each failure. Rather, it seems the httpd daemon hammers pam_unix via 'auth' too fast, and it results in the kernel log spitting out the following:

Dec 31 17:05:04 DD-WRT httpd[547]: pam_unix(httpd:auth): auth could not identify password for [root]
Dec 31 17:05:07 DD-WRT : Caught SIGABRT (6) sent by tkill 
Dec 31 17:05:07 DD-WRT : Thread 1097: httpd 
Dec 31 17:05:07 DD-WRT : === Context: 
Dec 31 17:05:07 DD-WRT : ZERO:00000000 AT:00000001 V0:00000000 V1:77799c38 A0:00000003 A1:77799c38 
Dec 31 17:05:07 DD-WRT : A2:00000000 A3:00000000 T0:00000000 T1:00000401 T2:00000051 T3:00001a1d 
Dec 31 17:05:07 DD-WRT : T4:00000000 T5:00002000 T6:00100000 T7:00001072 S0:77799c38 S1:00000000 
Dec 31 17:05:07 DD-WRT : S2:77b9f000 S3:77fb4000 S4:77a90b60 S5:00000002 S6:77799e28 S7:00000002 
Dec 31 17:05:07 DD-WRT : T8:00000000 T9:77a31780 K0:00000010 K1:00000000 GP:77ba3e20 SP:77799bb0 
Dec 31 17:05:07 DD-WRT : FP:77799e30 RA:77a18ce4 
Dec 31 17:05:07 DD-WRT : === Backtrace: 
Dec 31 17:05:07 DD-WRT : # [0x77a31790]: stack size 272 
Dec 31 17:05:07 DD-WRT : # [0x77a18ba4]: ra offset 316 
Dec 31 17:05:07 DD-WRT : # [0x77a18b9c]: stack size 320 
Dec 31 17:05:07 DD-WRT : # [0x77a79cc8]: stack size 96 
Dec 31 17:05:07 DD-WRT : # [0x77a79bec]: stack size 16 
Dec 31 17:05:07 DD-WRT : # [0x77a79b80]: stack size 16 
Dec 31 17:05:07 DD-WRT : # [0x77a79ad0]: ra offset 92 
Dec 31 17:05:07 DD-WRT : # [0x77a8e95c]: stack size 80 
Dec 31 17:05:07 DD-WRT : # [0x77a8e8d0]: ra offset 76 
Dec 31 17:05:07 DD-WRT : # [0x77a8e95c]: stack size 80 
Dec 31 17:05:07 DD-WRT : # [0x77a8e8d0]: ra offset 76 
Dec 31 17:05:07 DD-WRT : # [0x77a85008]: stack size 96 
Dec 31 17:05:07 DD-WRT : # [0x77a84e90]: ra offset 92 
Dec 31 17:05:07 DD-WRT : # [0x77a898fc]: ra offset 44 
Dec 31 17:05:07 DD-WRT : # [0x77a898e8]: stack size 48 
Dec 31 17:05:07 DD-WRT : # [0x77ba6f44]: ra offset 172 
Dec 31 17:05:07 DD-WRT : # [0x77ba6f18]: stack size 176 
Dec 31 17:05:07 DD-WRT : # [0x77b036fc]: stack size 32 
Dec 31 17:05:07 DD-WRT : # [0x77b036e8]: stack size 32 
Dec 31 17:05:07 DD-WRT : # [0x77b0368c]: stack size 32 
Dec 31 17:05:07 DD-WRT : /lib/libc.so.6[0x779ff000](gsignal+0x000000d0)[0x77a31850] 
Dec 31 17:05:07 DD-WRT : /lib/libc.so.6[0x779ff000](abort+0x00000154)[0x77a18ce4] 
Dec 31 17:05:07 DD-WRT : /lib/libc.so.6[0x779ff000](+0x0007adf8)[0x77a79df8] 
Dec 31 17:05:07 DD-WRT : /lib/libc.so.6[0x779ff000](strlen+0x00000000)[0x77a8ead0] (called 2 times) 
Dec 31 17:05:07 DD-WRT : /lib/libc.so.6[0x779ff000](+0x00086654)[0x77a85654] 
Dec 31 17:05:07 DD-WRT : /lib/libc.so.6[0x779ff000](+0x0008a954)[0x77a89954] 
Dec 31 17:05:07 DD-WRT : /lib/libpthread.so.0[0x77ba0000](+0x00007074)[0x77ba7074] 
Dec 31 17:05:07 DD-WRT : /lib/libc.so.6[0x779ff000](+0x0010471c)[0x77b0371c] 
Dec 31 17:05:07 DD-WRT : ???(+0)[0x00000000] 
Dec 31 17:05:07 DD-WRT : === Code: 
Dec 31 17:05:07 DD-WRT : 77a31810: 0000000c 00402025 2402107e 0000000c 00402825 02203025 240210aa 0000000c 
Dec 31 17:05:07 DD-WRT : 77a31830: 14e0000c 00408825 24040003 02002825 00003025 24070010 24021063 0000000c 
Dec 31 17:05:07 DD-WRT : 77a31850: >02201025 8fb00108 8fb1010c 03e00008 27bd0110 8f8294f8 7c03e83b 00431021 
Dec 31 17:05:07 DD-WRT : 77a31870: ac510000 1000fff0 2411ffff 00000000 3c1c0017 279c25a0 0399e021 04800005 

Note: say I change the password via 'passwd' in command line, the httpd daemon will pick it up in a matter of seconds and ask the user to re-authenticate before they are allowed to access the HTTPD UI (a good thing).

This makes me think that the drawing of 18-20 SVG graphics (in real time, one for each network interface; each of thse graphics have multiple time points to show network activity) is overwhelming the pam_unix authentication interface, resulting in it saying "could not identify", which then crashes the httpd daemon (which, of course, relies on this authentication) because of the volume of attempts.