Description Jari Kirma 2009-02-16 21:40:02 UTC

System (four-core Intel Q6600 with SMP kernel) crashes under load
(although relatively light load, thanks to only 2 Mbit outbound link) of
roughty hundred TCP connections somewhat reproducibly when HFSC ALTQ
traffic scheduling is used.

Only information seen about this is the dmesg message:

Fatal trap 12: page fault while in kernel mode
cpuid = 0; apic id = 00
fault virtual address   = 0x4
fault code              = supervisor read, page not present
instruction pointer     = 0x20:0xc04641e7
stack pointer           = 0x28:0xe719ca68
frame pointer           = 0x28:0xe719caac
code segment            = base rx0, limit 0xfffff, type 0x1b
                        = DPL 0, pres 1, def32 1, gran 1
processor eflags        = interrupt enabled, resume, IOPL = 0
current process         = 15 (swi4: clock sio)
trap number             = 12
panic: page fault

Where the instruction pointer points to inlined code inside hfsc_dequeue:

(gdb) l *(0xc04641e7)
0xc04641e7 is in hfsc_dequeue (altq_classq.h:113).
108             struct mbuf  *m, *m0;
109
110             if ((m = qtail(q)) == NULL)
111                     return (NULL);
112             if ((m0 = m->m_nextpkt) != m)
113                     m->m_nextpkt = m0->m_nextpkt;
114             else
115                     qtail(q) = NULL;
116             qlen(q)--;
117             m0->m_nextpkt = NULL;

Could it be just a simple locking issue in the linked list?

How-To-Repeat: Exact conditions are not known, but running lots of outbound TCP traffic over HFSC connection on a SMP system might trigger it.