VM Memory question

Having posted this to the wrong place.. :-0

Let me redirect it to the right one :-)

Randall Stewart wrote:
> Hi all:
> 
> I am looking at a memory problem.. i.e. maybe a leak somewhere..
> 
> I wrote a script in kgdb to get malloc and zone info.
> Now I am not all that familiar with these structures
> but I think there may be an issue.. either that
> or I just don't understand what I am reading :-)
> 
> The big standout to me is:
> 
> zone:0xc145f1e0 pages: 14092 keg_free:    0 size:16384 alloc:
>                 12446364 frees:  12446297 fail:     0 biu:57720832
> --------------------------------------------
> 
>  From my limited understanding.. pages aka keg->uk_pages
> tells me I have a huge number of pages in the zone
> coming out to about 57Meg of memory.. wow!
> 
> 
> Now this zone is really:
> -----------------------------------------------
> $3 = (struct uma_zone *) 0xc145f1e0
> (kgdb) print *$3
> $4 = {
>   uz_name = 0xc090c41b "mbuf_jumbo_16k",
>   uz_lock = 0xc1461188,
>   uz_keg = 0xc1461180,
>   uz_link = {
>     le_next = 0x0,
>     le_prev = 0xc14611ac
>   },
>   uz_full_bucket = {
>     lh_first = 0xc6a64c48
>   },
>   uz_free_bucket = {
>     lh_first = 0xc6a3b624
>   },
>   uz_ctor = 0xc06888ec <mb_ctor_clust>,
>   uz_dtor = 0xc06889c4 <mb_dtor_clust>,
>   uz_init = 0,
>   uz_fini = 0,
>   uz_allocs = 0xbdea9c,
>   uz_frees = 0xbdea59,
>   uz_fails = 0x0,
>   uz_fills = 0x0,
>   uz_count = 0x80,
>   uz_cpu = {{
>       uc_freebucket = 0xc6a3f20c,
>       uc_allocbucket = 0xc6a35000,
>       uc_allocs = 0x291,
>       uc_frees = 0x295
>     }}
> }
> ------------------------------------
> 
> Now looking at the allocs/frees.. and subtracting we
> only have (according to the zone) 67 that are not free..
> 
> The keg looks as follows:
> ----------------------------------------
> $5 = {
>   uk_link = {
>     le_next = 0xc1461100,
>     le_prev = 0xc1461200
>   },
>   uk_lock = {
>     mtx_object = {
>       lo_name = 0xc090c41b "mbuf_jumbo_16k",
>       lo_type = 0xc09099e2 "UMA zone",
>       lo_flags = 0x1430000,
>       lo_witness_data = {
>         lod_list = {
>           stqe_next = 0xc0a06be0
>         },
>         lod_witness = 0xc0a06be0
>       }
>     },
>     mtx_lock = 0x4,
>     mtx_recurse = 0x0
>   },
>   uk_hash = {
>     uh_slab_hash = 0xc6a4b000,
>     uh_hashsize = 0x1000,
>     uh_hashmask = 0xfff
>   },
>   uk_zones = {
>     lh_first = 0xc145f1e0
>   },
>   uk_part_slab = {
>     lh_first = 0x0
>   },
>   uk_free_slab = {
>     lh_first = 0x0
>   },
>   uk_full_slab = {
>     lh_first = 0xc692b5b0
>   },
>   uk_recurse = 0x0,
>   uk_align = 0x3,
>   uk_pages = 0x370c,
>   uk_free = 0x0,
>   uk_size = 0x4000,
>   uk_rsize = 0x4000,
>   uk_maxpages = 0x0,
>   uk_init = 0xc0829b90 <trash_init>,
>   uk_fini = 0xc0829ba8 <trash_fini>,
>   uk_allocf = 0xc0826f40 <page_alloc>,
>   uk_freef = 0xc08270ac <page_free>,
>   uk_obj = 0x0,
>   uk_kva = 0x0,
>   uk_slabzone = 0xc14731e0,
>   uk_pgoff = 0x0,
>   uk_ppera = 0x4,
>   uk_ipers = 0x1,
>   uk_flags = 0x508
> }
> ----------------------------------
> 
> So where is all the memory hiding?
> 
> And if its free (as the zone implies) why is it
> not getting released back to the system... (the machine
> crashed with a "no memory" condition..
> 
> ???
> 
> Help from someone with a clue would be much
> appreciated :-)
> 
> Thanks
> 
> R


-- 
Randall Stewart
NSSTG - Cisco Systems Inc.
803-345-0369 <or> 803-317-4952 (cell)