alloc_large_system_hash() is used to allocate large hash tables at boot time.
Example on a 2 nodes NUMA machine :
Inode-cache hash table entries: 1048576 (order: 11, 8388608 bytes)
IP route cache hash table entries: 2097152 (order: 12, 16777216 bytes)
TCP established hash table entries: 2097152 (order: 12, 16777216 bytes)
Memory is taken from :
bootmem if (flags & HASH_EARLY)
__vmalloc() if (hashdist is set) (NUMA knob)
__get_free_pages(GFP_ATOMIC, order);
What would be the needed changes in the code to get both :
- Allocate ram equally from all the nodes of the machine
- Use large pages (2MB) to lower TLB stress
Thank you
Eric
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
- Follow-Ups:
- Re: Questions about alloc_large_system_hash() and TLB entries
- From: "David S. Miller" <[email protected]>
- Re: Questions about alloc_large_system_hash() and TLB entries
- References:
- [PATCH] i386: Add a temporary to make put_user more type safe.
- From: [email protected] (Eric W. Biederman)
- Re: [PATCH] i386: Add a temporary to make put_user more type safe.
- From: [email protected] (Eric W. Biederman)
- Re: [PATCH] i386: Add a temporary to make put_user more type safe.
- From: Andrew Morton <[email protected]>
- Re: [PATCH] i386: Add a temporary to make put_user more type safe.
- From: Andrew Morton <[email protected]>
- [PATCH] i386: instead of poisoning .init zone, change protection bits to force a fault
- From: Eric Dumazet <[email protected]>
- Re: [PATCH] i386: instead of poisoning .init zone, change protection bits to force a fault
- From: Benjamin LaHaise <[email protected]>
- [PATCH, V2] i386: instead of poisoning .init zone, change protection bits to force a fault
- From: Eric Dumazet <[email protected]>
- [PATCH] i386: Add a temporary to make put_user more type safe.
- Prev by Date: [patch] fix alarm() return value
- Next by Date: Re: [PATCH RFC] put UTS_RELEASE in separate utsversion.h file
- Previous by thread: [PATCH, V2] i386: instead of poisoning .init zone, change protection bits to force a fault
- Next by thread: Re: Questions about alloc_large_system_hash() and TLB entries
- Index(es):
 |