as opposed to the classic Dewey, Cheatham, and Howe

ok.. this is interesting.. even on the CX-5 cards, reading MLXCX_REG_PCAM failes with MLXCX_CMD_R_BAD_SYS_STATE

at least the ones in this system

w

oops

Anyone use or Check out github.com/FDio/vpp ?

Ahh nevermind it’s another kernel bypass thingie

is src.iluumos.org very busy or is it my network....

got 504 gateway timeout from it...

src.illumos.org even :)

it works here, but sometimes it stalls on people

(it works when I spell it right. Did you?)

:)

i better kick my router just in case...

so, mdb ::mutex command will print put info about mutex, the held pointer there is pointer to kthread_t? and to get the stack trace, I can just use that pointer with ::stacks ?

::stack (no s)

::stacks does something different

I usually use ::findstack -v

err yeah

so, UTM with cores > 1 usually will hung on boot. so I found thread with start_cpu() in its stack and wanted to check the mutex and found this paste.ec/paste/sG5kZxnU#rzagv9Op+ifRrqGXUfLVXgmP8HFksN9tAm8Cwa1D4c8 -- as much as I can see, cpus should be cpuset_t in data segment and swtch_from_zombie() should be busy with resume_from_zombie()

not really sure if that start_cpu() really is guilty one, but found this example to be interesting....

ofc course resume_from_zombie() does some things with stack...

::stack doesn't take a kthread_t

::findstack does, ::stacks also does I think

yes it does.

stacks and findstack agree there

::stack/$C take a frame pointer

does anyone know of a debugger that visualizes the stack and annotates it with the %pc when %sp was adjusted?

(read %rip/%rsp/%rbp if you'd like)

I haven't used a debugger that had anything approaching a gui since DDD in the 90s

that is the actual _stack_ not a stack trace. I'm basically asking for $C but with the data between each %fp ::dump'd

(it feels like with dwarf you could make that substantially less grody, which is why I'm asking)

have you checked this one: github.com/cyrus-and/gdb-dashboard -- altho I'm not sure it can do what you are asking for:D

gosh that's busy. It might help though!

I have found it useful sometimes....

hm, boot -B smt_enabled=false and no hung...

tested smt_enabled=false because of this: paste.ec/paste/Fy63rJu4#VMxs9xHP49Om+8iMZbH-ZJwuTD6Nch1T8Dp1TKGEjCk

hm, apparently no, it was just one of random good boot..