Hi All,
I have several questions about swapped out kernel memory:
1. What kind of kernel memory can be swapped out?
As I known before, kernel heap is a nonpageable region, which is
management by segkmem. Nonpageable means that it doesn't allow pagefault
and swaped out. And kthread_t's stack came from segkp, which is pageable
kernel memory region.
Does it means the all memory managed by segkp can be swaped out?
Beside this, any other portions of kernel memory can be swaped out?
2. How kthread's stack swaped out can impact the findleaks results?
Somtimes, in ::findleaks 's output, I had seen many kthread's stack was
swaped out like this:
> ::findleaks -d
findleaks: thread ffffffff82af7720
's stack swapped out; false positives possible
findleaks: thread ffffffff82af6c40
's stack swapped out; false positives possible
..............................
.................................
...........................................
> ffffffff82af7720::print kthread_t t_pcb.val[1]
t_pcb.val[1] = 0xfffffe80003a2c50
> 0xfffffe80003a2c50::whatis
fffffe80003a2c50 is in thread ffffffff82af7720's stack
> 0xfffffe80003a2c50,30/nap
0xfffffe80003a2c50:
mdb: failed to read data from target: no mapping for address
0xfffffe80003a2c50:
mdb: failed to read data from target: no mapping for address
0xfffffe80003a2c50:
...............................................
..............................................
found leaks here....
................................
3. Could we trust the leaks found by ::findleaks in above case?
When lots of kthread's stack swaped out was found, it indicates mdb
couldn't verify the all of memory reference in kthread's stack, maybe
leaked memory's reference exists in those swaped out stacks.
If it's true, we couldn't trust the kmem leaks reported by ::findleaks
in this case, right?
--
Cheers,
----------------------------------------------------------------------
Oliver Yang | OPG Engineering Operation | [EMAIL PROTECTED] | x82229
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
