I both agree and disagree with Jeff. I think the only *disadvantage*
to using objcache for struct proc is that objcache must cache a minimum
number of free structures to be efficient so the overall kernel memory
use would be a bit greater verses uses kmalloc(). But I agree with Jeff
in that objcache's biggest advantage is that you can keep structural state
intact and struct proc might not be a good candidate (yet).
What I suggest is that you convert struct proc from zone to kmalloc
first, and then look into a possible objcache adaptation. kmalloc
is not slow.. it's a very fast allocator. Objcache is probably a tad
faster but not by much. Where objcache really wins is in not having
to completey reinitialize the allocated structures.
Over the years I've removed (I think) all the dependancies on requiring
the proc structure to be in stable memory so I believe it can be moved to
kmalloc/objcache now without any major pain.
-Matt
Matthew Dillon
<dillon@backplane.com>
| Ryan Hope | reiser4 for 2.6.27-rc1 |
| Tim Tassonis | reiser4 for 2.6.27-rc1 |
| Rafael J. Wysocki | 2.6.26-rc9-git12: Reported regressions from 2.6.25 |
| hooanon05 | [PATCH 63/67] aufs mount helper |
git: | |
| Petko Manolov | git and binary files |
| Wink Saville | Resolving conflicts |
| Ken Pratt | pack operation is thrashing my server |
| Jeff King | Re: Use of tac in git-rebase--interactive |
| GVG GVG | ssh_exchange_identification: Connection closed by remote host |
| Julien TOUCHE | setting up ssh tunnel/vpn |
| Ray Percival | Re: Real men don't attack straw men |
| Michael | Performance: OpenVPN vs IPsec |
| Pietro Cerutti | Re: [patch] enhance powerd(8) to handle max temperature |
| Søren Schmidt | Re: Patch RFC: Promise SATA300 TX4 hardware bug workaround. |
| Kris Kennaway | Re: Problems with FreeBSD 6.0 |
| Ceri Davies | Using any network interface whatsoever |
