[Devel] Re: [PATCH 6/7] BC: kernel memory (core)

[Kirill Korotaev]

Balbir Singh wrote:
> Kirill Korotaev wrote:
> 
>> Introduce BC_KMEMSIZE resource which accounts kernel
>> objects allocated by task's request.
>>
>> Reference to BC is kept on struct page or slab object.
>> For slabs each struct slab contains a set of pointers
>> corresponding objects are charged to.
>>
>> Allocation charge rules:
>> 1. Pages - if allocation is performed with __GFP_BC flag - page
>>    is charged to current's exec_bc.
>> 2. Slabs - kmem_cache may be created with SLAB_BC flag - in this
>>    case each allocation is charged. Caches used by kmalloc are
>>    created with SLAB_BC | SLAB_BC_NOCHARGE flags. In this case
>>    only __GFP_BC allocations are charged.
>>
> 
> <snip>
> 
>> +#define __GFP_BC_LIMIT ((__force gfp_t)0x100000u) /* Charge against 
>> BC limit */
>>
> 
> What's _GFP_BC_LIMIT for, could you add the description for that flag?
> The comment is not very clear
> 
>> +#ifdef CONFIG_BEANCOUNTERS
>> +    union {
>> +        struct beancounter    *page_bc;
>> +    } bc;
>> +#endif
>> };
>>
>> +#define page_bc(page)            ((page)->bc.page_bc)
> 
> 
> Minor comment - page->(bc).page_bc has too many repititions of page and 
> bc - see
> the Practice of Programming by Kernighan and Pike
> 
> I missed the part of why you wanted to have a union (in struct page for 
> bc)?
because this union is used both for kernel memory accounting and user memeory tracking.

>> const char *bc_rnames[] = {
>> +    "kmemsize",    /* 0 */
>> };
>>
>> static struct hlist_head bc_hash[BC_HASH_SIZE];
>> @@ -221,6 +222,8 @@ static void init_beancounter_syslimits(s
>> {     int k;
>>
>> +    bc->bc_parms[BC_KMEMSIZE].limit = 32 * 1024 * 1024;
>> +
> 
> 
> Can't this be configurable CONFIG_XXX or a #defined constant?
This is some arbitraty limited container, just to make sure it is not
created unlimited. User space should initialize limits properly after creation
anyway. So I don't see reasons to make it configurable, do you?

>> --- ./mm/mempool.c.bckmem    2006-04-21 11:59:36.000000000 +0400
>> +++ ./mm/mempool.c    2006-08-28 12:59:28.000000000 +0400
>> @@ -119,6 +119,7 @@ int mempool_resize(mempool_t *pool, int     
>> unsigned long flags;
>>
>>     BUG_ON(new_min_nr <= 0);
>> +    gfp_mask &= ~__GFP_BC;
>>
>>     spin_lock_irqsave(&pool->lock, flags);
>>     if (new_min_nr <= pool->min_nr) {
>> @@ -212,6 +213,7 @@ void * mempool_alloc(mempool_t *pool, gf
>>     gfp_mask |= __GFP_NOMEMALLOC;    /* don't allocate emergency 
>> reserves */
>>     gfp_mask |= __GFP_NORETRY;    /* don't loop in __alloc_pages */
>>     gfp_mask |= __GFP_NOWARN;    /* failures are OK */
>> +    gfp_mask &= ~__GFP_BC;        /* do not charge */
>>
>>     gfp_temp = gfp_mask & ~(__GFP_WAIT|__GFP_IO);
>>
> 
> Is there any reasn why mempool_xxxx() functions are not charged? Is it 
> because
> mempool functions are mostly used from the I/O path?
yep.

>> --- ./mm/page_alloc.c.bckmem    2006-08-28 12:20:13.000000000 +0400
>> +++ ./mm/page_alloc.c    2006-08-28 12:59:28.000000000 +0400
>> @@ -40,6 +40,8 @@
>> #include <linux/sort.h>
>> #include <linux/pfn.h>
>>
>> +#include <bc/kmem.h>
>> +
>> #include <asm/tlbflush.h>
>> #include <asm/div64.h>
>> #include "internal.h"
>> @@ -516,6 +518,8 @@ static void __free_pages_ok(struct page     if 
>> (reserved)
>>         return;
>>
>> +    bc_page_uncharge(page, order);
>> +
>>     kernel_map_pages(page, 1 << order, 0);
>>     local_irq_save(flags);
>>     __count_vm_events(PGFREE, 1 << order);
>> @@ -799,6 +803,8 @@ static void fastcall free_hot_cold_page(
>>     if (free_pages_check(page))
>>         return;
>>
>> +    bc_page_uncharge(page, 0);
>> +
>>     kernel_map_pages(page, 1, 0);
>>
>>     pcp = &zone_pcp(zone, get_cpu())->pcp[cold];
>> @@ -1188,6 +1194,11 @@ nopage:
>>         show_mem();
>>     }
>> got_pg:
>> +    if ((gfp_mask & __GFP_BC) &&
>> +            bc_page_charge(page, order, gfp_mask)) {
> 
> 
> I wonder if bc_page_charge() should be called bc_page_charge_failed()?
> Does it make sense to atleast partially start reclamation here? I know with
> bean counters we cannot reclaim from a particular container, but for now
> we could kick off kswapd() or call shrink_all_memory() inline (Dave's 
> patches do this to shrink memory from the particular cpuset). Or do you 
> want to leave this
> slot open for later?
yes. my intention is to account correctly all needed information first.
After we agree on accounting, we can agree on how to do reclamaition.

>> +        __free_pages(page, order);
>> +        page = NULL;
>> +    }
> 
> 
>