fxp unaligned access

[Andrew Gallatin]

Fred Clift writes:
 > On Fri, 25 Apr 2003, Maxime Henrion wrote:
 > 
 > >
 > > Can you, in gdb, go to the fxp_add_rfabuf() frame (use the frame #
 > > command) and try this :
 > >
 > > p sc
 > > p rfa

 > under the circumstances -- but I'll try :)  The other option would be to
 > build without fxp in the kernel and see if loading the module has the same
 > effect.
 > 

Don't bother to try to get a crashdump, it won't work on a trap
anyway.  It will only work on an explicit panic.

BTW, from your original message:


 ) fatal kernel trap:
 ) 
 )     trap entry     = 0x4 (unaligned access fault)
 )     faulting va    = 0xfffffe00004b780e
 )     opcode         = 0x28
 )     register       = 0x1
 )     pc             = 0xfffffc00003bb180
 )     ra             = 0xfffffc00003bb13c
 )     sp             = 0xfffffc00008639a0
 )     usp            = 0x0
 )     curthread      = 0xfffffc0000725ea8
 )         pid = 0, comm = swapper
 ) 
 ) Stopped at      fxp_add_rfabuf+0x80:    zapnot  t0,#0xf3,t0     <t0=0x10>
 ) db> t
 ) fxp_add_rfabuf() at fxp_add_rfabuf+0x80
 ) fxp_attach() at fxp_attach+0xd7c
 ) device_probe_and_attach() at device_probe_and_attach+0xc8
 ) bus_generic_attach() at bus_generic_attach+0x28
 ) <..>

Isn't htole16 just supposed to be a cast to (u_int16_t) on alpha?

It might be interesting to try just using the cast..


Drew