Messages in this thread | | | Date | Fri, 28 Jul 2023 17:35:33 +0100 | From | James Morse <> | Subject | Re: [PATCH v4 06/24] x86/resctrl: Track the number of dirty RMID a CLOSID has |
| |
Hi Reinette,
On 6/15/23 23:08, Reinette Chatre wrote: > On 5/25/2023 11:01 AM, James Morse wrote: >> @@ -420,6 +438,9 @@ static void add_rmid_to_limbo(struct rmid_entry *entry) >> rmid_limbo_count++; >> else >> list_add_tail(&entry->list, &rmid_free_lru); >> + >> + if (IS_ENABLED(CONFIG_RESCTRL_RMID_DEPENDS_ON_CLOSID)) >> + closid_num_dirty_rmid[entry->closid]++;
> Wouldn't this always increment the counter, whether the entry is > dirty or not? (Although ... the later change where entries are > always dirty may make this correct ... although I would still > expect the if statement that precedes it to change).
True, I was expecting add_rmid_to_limbo() to always transiently add CLOSID to limbo, hence this is unconditional, but you're right its optional - and this could cause everything to pile up in the list when the limbo handler isn't running. I'll add a check on entry->busy, and move this into the above if case.
Thanks,
James
| |