From: Vladislav B. <vs...@vl...> - 2006-01-25 18:22:26
|
Nathaniel Clark wrote: > Here is an updated version of my target driver + patch for 2.6 kernels Good work. > Features over my first drop: > * this one works > * disabling LUNs isn't brain dead and will actually disable the LUN > * tested and seems relatively stable > * uses slab alloc/free instead of kmalloc/kfree for commands (saves > memory & time, and is big red flag for leaky packets) Nice. I always wanted to do it and always had no time (or forgot) :) > * paths through code are somewhat simplified (mostly in the various > (de)init/(un)register paths) > * transfers of large payloads works 16k+ (old code would loose CTIO > packet if it used a continuation packet) Strange, how then it currently works? 16k+ are quite usual. Could you be more prrrrecise, please? Some comments/questions after a brief review (I'll review it closer when I finish with SCST sessions registration rework. Then I'll convert it to the new interface and send you back. Well, looks like we starting need CVS on sf.net..). - Why do you added in q2t_target_release() prepare_to_wait() etc. Isn't schedule_timeout() not enough (as it was)? - You forgot to add RC_MGMT definition in the Makefile and require (#error) it in qla2x00t.c. BTW, why do you think it should be required? Vlad |