[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-devel] [PATCH 6 of 6] x86/mm: Don't claim a slot on the paging ring if we might not need it
# HG changeset patch # User Tim Deegan <tim@xxxxxxx> # Date 1330013929 0 # Node ID 516f96172e67ff463c371357e0ab93724c8b9048 # Parent 510d80343793227bd39b9a5d4d3b61b9a3f776e4 x86/mm: Don't claim a slot on the paging ring if we might not need it. This avoids the possibility of hitting a wait queue for a slot and then finding out later that it wasn't necessary. Signed-off-by: Tim Deegan <tim@xxxxxxx> diff -r 510d80343793 -r 516f96172e67 xen/arch/x86/mm/p2m.c --- a/xen/arch/x86/mm/p2m.c Thu Feb 23 16:18:09 2012 +0000 +++ b/xen/arch/x86/mm/p2m.c Thu Feb 23 16:18:49 2012 +0000 @@ -1001,19 +1001,23 @@ void p2m_mem_paging_populate(struct doma p2m_access_t a; mfn_t mfn; struct p2m_domain *p2m = p2m_get_hostp2m(d); - int send_request = 0; + int rc, send_request = 0; - /* We're paging. There should be a ring */ - int rc = mem_event_claim_slot(d, &d->mem_event->paging); - if ( rc == -ENOSYS ) + if ( !d->mem_event->paging.ring_page ) { gdprintk(XENLOG_ERR, "Domain %hu paging gfn %lx yet no ring " "in place\n", d->domain_id, gfn); domain_crash(d); return; } - else if ( rc < 0 ) - return; + + /* Foreign users must grab a ring entry before doing any work since + * they might not be able to get one later. Local users shouldn't + * grab a slot until they know they need it, to avoid going on a + * wait-queue unnecessarily. */ + if ( d != current->domain ) + if ( mem_event_claim_slot(d, &d->mem_event->paging) < 0 ) + return; /* Fix p2m mapping */ gfn_lock(p2m, gfn, 0); @@ -1042,10 +1046,18 @@ void p2m_mem_paging_populate(struct doma if ( p2mt != p2m_ram_paging_out && p2mt != p2m_ram_paged ) { /* gfn is already on its way back and vcpu is not paused */ - mem_event_cancel_slot(d, &d->mem_event->paging); + if ( d != current->domain ) + mem_event_cancel_slot(d, &d->mem_event->paging); return; } + /* Now local users should claim a slot. */ + if ( d == current->domain ) + { + rc = mem_event_claim_slot(d, &d->mem_event->paging); + ASSERT(rc == 0); + } + /* Send request to pager */ req.p2mt = p2mt; req.vcpu_id = v->vcpu_id; _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxx http://lists.xen.org/xen-devel
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |