From: Avi Kivity <avi@redhat.com> To: Liu Ping Fan <qemulist@gmail.com> Cc: kvm@vger.kernel.org, "Jan Kiszka" <jan.kiszka@siemens.com>, "Marcelo Tosatti" <mtosatti@redhat.com>, qemu-devel@nongnu.org, "Blue Swirl" <blauwirbel@gmail.com>, "Anthony Liguori" <anthony@codemonkey.ws>, "Stefan Hajnoczi" <stefanha@gmail.com>, "Paolo Bonzini" <pbonzini@redhat.com>, "Andreas Färber" <afaerber@suse.de> Subject: Re: [PATCH 09/15] memory: prepare flatview and radix-tree for rcu style access Date: Wed, 08 Aug 2012 12:41:03 +0300 [thread overview] Message-ID: <5022342F.2070609@redhat.com> (raw) In-Reply-To: <1344407156-25562-10-git-send-email-qemulist@gmail.com> On 08/08/2012 09:25 AM, Liu Ping Fan wrote: > From: Liu Ping Fan <pingfank@linux.vnet.ibm.com> > > Flatview and radix view are all under the protection of pointer. > And this make sure the change of them seem to be atomic! > > The mr accessed by radix-tree leaf or flatview will be reclaimed > after the prev PhysMap not in use any longer > IMO this cleverness should come much later. Let's first take care of dropping the big qemu lock, then make swithcing memory maps more efficient. The initial paths could look like: lookup: take mem_map_lock lookup take ref drop mem_map_lock update: take mem_map_lock (in core_begin) do updates drop memo_map_lock Later we can replace mem_map_lock with either a rwlock or (real) rcu. > > #if !defined(CONFIG_USER_ONLY) > > -static void phys_map_node_reserve(unsigned nodes) > +static void phys_map_node_reserve(PhysMap *map, unsigned nodes) > { > - if (phys_map_nodes_nb + nodes > phys_map_nodes_nb_alloc) { > + if (map->phys_map_nodes_nb + nodes > map->phys_map_nodes_nb_alloc) { > typedef PhysPageEntry Node[L2_SIZE]; > - phys_map_nodes_nb_alloc = MAX(phys_map_nodes_nb_alloc * 2, 16); > - phys_map_nodes_nb_alloc = MAX(phys_map_nodes_nb_alloc, > - phys_map_nodes_nb + nodes); > - phys_map_nodes = g_renew(Node, phys_map_nodes, > - phys_map_nodes_nb_alloc); > + map->phys_map_nodes_nb_alloc = MAX(map->phys_map_nodes_nb_alloc * 2, > + 16); > + map->phys_map_nodes_nb_alloc = MAX(map->phys_map_nodes_nb_alloc, > + map->phys_map_nodes_nb + nodes); > + map->phys_map_nodes = g_renew(Node, map->phys_map_nodes, > + map->phys_map_nodes_nb_alloc); > } > } Please have a patch that just adds the map parameter to all these functions. This makes the later patch, that adds the copy, easier to read. > + > +void cur_map_update(PhysMap *next) > +{ > + qemu_mutex_lock(&cur_map_lock); > + physmap_put(cur_map); > + cur_map = next; > + smp_mb(); > + qemu_mutex_unlock(&cur_map_lock); > +} IMO this can be mem_map_lock. If we take my previous suggestion: lookup: take mem_map_lock lookup take ref drop mem_map_lock update: take mem_map_lock (in core_begin) do updates drop memo_map_lock And update it to update: prepare next_map (in core_begin) do updates take mem_map_lock (in core_commit) switch maps drop mem_map_lock free old map Note the lookup path copies the MemoryRegionSection instead of referencing it. Thus we can destroy the old map without worrying; the only pointers will point to MemoryRegions, which will be protected by the refcounts on their Objects. This can be easily switched to rcu: update: prepare next_map (in core_begin) do updates switch maps - rcu_assign_pointer call_rcu(free old map) (or synchronize_rcu; free old maps) Again, this should be done after the simplictic patch that enables parallel lookup but keeps just one map. -- error compiling committee.c: too many arguments to function
WARNING: multiple messages have this Message-ID (diff)
From: Avi Kivity <avi@redhat.com> To: Liu Ping Fan <qemulist@gmail.com> Cc: kvm@vger.kernel.org, "Jan Kiszka" <jan.kiszka@siemens.com>, "Marcelo Tosatti" <mtosatti@redhat.com>, qemu-devel@nongnu.org, "Blue Swirl" <blauwirbel@gmail.com>, "Anthony Liguori" <anthony@codemonkey.ws>, "Stefan Hajnoczi" <stefanha@gmail.com>, "Paolo Bonzini" <pbonzini@redhat.com>, "Andreas Färber" <afaerber@suse.de> Subject: Re: [Qemu-devel] [PATCH 09/15] memory: prepare flatview and radix-tree for rcu style access Date: Wed, 08 Aug 2012 12:41:03 +0300 [thread overview] Message-ID: <5022342F.2070609@redhat.com> (raw) In-Reply-To: <1344407156-25562-10-git-send-email-qemulist@gmail.com> On 08/08/2012 09:25 AM, Liu Ping Fan wrote: > From: Liu Ping Fan <pingfank@linux.vnet.ibm.com> > > Flatview and radix view are all under the protection of pointer. > And this make sure the change of them seem to be atomic! > > The mr accessed by radix-tree leaf or flatview will be reclaimed > after the prev PhysMap not in use any longer > IMO this cleverness should come much later. Let's first take care of dropping the big qemu lock, then make swithcing memory maps more efficient. The initial paths could look like: lookup: take mem_map_lock lookup take ref drop mem_map_lock update: take mem_map_lock (in core_begin) do updates drop memo_map_lock Later we can replace mem_map_lock with either a rwlock or (real) rcu. > > #if !defined(CONFIG_USER_ONLY) > > -static void phys_map_node_reserve(unsigned nodes) > +static void phys_map_node_reserve(PhysMap *map, unsigned nodes) > { > - if (phys_map_nodes_nb + nodes > phys_map_nodes_nb_alloc) { > + if (map->phys_map_nodes_nb + nodes > map->phys_map_nodes_nb_alloc) { > typedef PhysPageEntry Node[L2_SIZE]; > - phys_map_nodes_nb_alloc = MAX(phys_map_nodes_nb_alloc * 2, 16); > - phys_map_nodes_nb_alloc = MAX(phys_map_nodes_nb_alloc, > - phys_map_nodes_nb + nodes); > - phys_map_nodes = g_renew(Node, phys_map_nodes, > - phys_map_nodes_nb_alloc); > + map->phys_map_nodes_nb_alloc = MAX(map->phys_map_nodes_nb_alloc * 2, > + 16); > + map->phys_map_nodes_nb_alloc = MAX(map->phys_map_nodes_nb_alloc, > + map->phys_map_nodes_nb + nodes); > + map->phys_map_nodes = g_renew(Node, map->phys_map_nodes, > + map->phys_map_nodes_nb_alloc); > } > } Please have a patch that just adds the map parameter to all these functions. This makes the later patch, that adds the copy, easier to read. > + > +void cur_map_update(PhysMap *next) > +{ > + qemu_mutex_lock(&cur_map_lock); > + physmap_put(cur_map); > + cur_map = next; > + smp_mb(); > + qemu_mutex_unlock(&cur_map_lock); > +} IMO this can be mem_map_lock. If we take my previous suggestion: lookup: take mem_map_lock lookup take ref drop mem_map_lock update: take mem_map_lock (in core_begin) do updates drop memo_map_lock And update it to update: prepare next_map (in core_begin) do updates take mem_map_lock (in core_commit) switch maps drop mem_map_lock free old map Note the lookup path copies the MemoryRegionSection instead of referencing it. Thus we can destroy the old map without worrying; the only pointers will point to MemoryRegions, which will be protected by the refcounts on their Objects. This can be easily switched to rcu: update: prepare next_map (in core_begin) do updates switch maps - rcu_assign_pointer call_rcu(free old map) (or synchronize_rcu; free old maps) Again, this should be done after the simplictic patch that enables parallel lookup but keeps just one map. -- error compiling committee.c: too many arguments to function
next prev parent reply other threads:[~2012-08-08 9:41 UTC|newest] Thread overview: 154+ messages / expand[flat|nested] mbox.gz Atom feed top 2012-08-08 6:25 [PATCH 0/15 v2] prepare unplug out of protection of global lock Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 6:25 ` [PATCH 01/15] atomic: introduce atomic operations Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 8:55 ` Paolo Bonzini 2012-08-08 8:55 ` [Qemu-devel] " Paolo Bonzini 2012-08-08 9:02 ` Avi Kivity 2012-08-08 9:02 ` [Qemu-devel] " Avi Kivity 2012-08-08 9:05 ` 陳韋任 (Wei-Ren Chen) 2012-08-08 9:05 ` 陳韋任 (Wei-Ren Chen) 2012-08-08 9:15 ` Avi Kivity 2012-08-08 9:15 ` [Qemu-devel] " Avi Kivity 2012-08-08 9:21 ` Peter Maydell 2012-08-08 9:21 ` Peter Maydell 2012-08-08 13:09 ` Stefan Hajnoczi 2012-08-08 13:09 ` Stefan Hajnoczi 2012-08-08 13:18 ` Paolo Bonzini 2012-08-08 13:18 ` Paolo Bonzini 2012-08-08 13:32 ` Peter Maydell 2012-08-08 13:32 ` [Qemu-devel] " Peter Maydell 2012-08-08 13:49 ` Paolo Bonzini 2012-08-08 13:49 ` [Qemu-devel] " Paolo Bonzini 2012-08-08 14:00 ` Avi Kivity 2012-08-08 14:00 ` [Qemu-devel] " Avi Kivity 2012-08-08 6:25 ` [PATCH 02/15] qom: using atomic ops to re-implement object_ref Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 6:25 ` [PATCH 03/15] qom: introduce reclaimer to release obj Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:05 ` Avi Kivity 2012-08-08 9:05 ` [Qemu-devel] " Avi Kivity 2012-08-08 9:07 ` Paolo Bonzini 2012-08-08 9:07 ` [Qemu-devel] " Paolo Bonzini 2012-08-08 9:15 ` Avi Kivity 2012-08-08 9:15 ` [Qemu-devel] " Avi Kivity 2012-08-09 7:33 ` liu ping fan 2012-08-09 7:33 ` [Qemu-devel] " liu ping fan 2012-08-09 7:49 ` Paolo Bonzini 2012-08-09 7:49 ` [Qemu-devel] " Paolo Bonzini 2012-08-09 8:18 ` Avi Kivity 2012-08-09 8:18 ` [Qemu-devel] " Avi Kivity 2012-08-10 6:43 ` liu ping fan 2012-08-10 6:43 ` [Qemu-devel] " liu ping fan 2012-08-08 9:35 ` Paolo Bonzini 2012-08-08 9:35 ` [Qemu-devel] " Paolo Bonzini 2012-08-09 7:38 ` liu ping fan 2012-08-09 7:38 ` [Qemu-devel] " liu ping fan 2012-08-08 6:25 ` [PATCH 04/15] memory: MemoryRegion topology must be stable when updating Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:13 ` Avi Kivity 2012-08-08 9:13 ` [Qemu-devel] " Avi Kivity 2012-08-09 7:28 ` liu ping fan 2012-08-09 7:28 ` [Qemu-devel] " liu ping fan 2012-08-09 8:24 ` Avi Kivity 2012-08-09 8:24 ` [Qemu-devel] " Avi Kivity 2012-08-10 6:44 ` liu ping fan 2012-08-10 6:44 ` [Qemu-devel] " liu ping fan 2012-08-13 18:28 ` Marcelo Tosatti 2012-08-13 18:28 ` [Qemu-devel] " Marcelo Tosatti 2012-08-08 19:17 ` Blue Swirl 2012-08-08 19:17 ` [Qemu-devel] " Blue Swirl 2012-08-09 7:28 ` liu ping fan 2012-08-09 7:28 ` [Qemu-devel] " liu ping fan 2012-08-09 17:09 ` Blue Swirl 2012-08-09 17:09 ` [Qemu-devel] " Blue Swirl 2012-08-08 6:25 ` [PATCH 05/15] memory: introduce life_ops to MemoryRegion Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:18 ` Avi Kivity 2012-08-08 9:18 ` [Qemu-devel] " Avi Kivity 2012-08-08 6:25 ` [PATCH 06/15] memory: use refcnt to manage MemoryRegion Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:20 ` Avi Kivity 2012-08-08 9:20 ` [Qemu-devel] " Avi Kivity 2012-08-09 7:27 ` liu ping fan 2012-08-09 7:27 ` [Qemu-devel] " liu ping fan 2012-08-09 8:38 ` Avi Kivity 2012-08-09 8:38 ` [Qemu-devel] " Avi Kivity 2012-08-10 6:44 ` liu ping fan 2012-08-10 6:44 ` [Qemu-devel] " liu ping fan 2012-08-12 8:43 ` Avi Kivity 2012-08-12 8:43 ` [Qemu-devel] " Avi Kivity 2012-08-08 6:25 ` [PATCH 07/15] memory: inc/dec mr's ref when adding/removing from mem view Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 6:25 ` [PATCH 08/15] memory: introduce PhysMap to present snapshot of toploygy Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:27 ` Avi Kivity 2012-08-08 9:27 ` [Qemu-devel] " Avi Kivity 2012-08-08 19:18 ` Blue Swirl 2012-08-08 19:18 ` [Qemu-devel] " Blue Swirl 2012-08-09 7:29 ` liu ping fan 2012-08-09 7:29 ` [Qemu-devel] " liu ping fan 2012-08-08 6:25 ` [PATCH 09/15] memory: prepare flatview and radix-tree for rcu style access Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:41 ` Avi Kivity [this message] 2012-08-08 9:41 ` Avi Kivity 2012-08-11 1:58 ` liu ping fan 2012-08-11 1:58 ` [Qemu-devel] " liu ping fan 2012-08-11 10:06 ` liu ping fan 2012-08-11 10:06 ` [Qemu-devel] " liu ping fan 2012-08-08 19:23 ` Blue Swirl 2012-08-08 19:23 ` [Qemu-devel] " Blue Swirl 2012-08-09 7:29 ` liu ping fan 2012-08-09 7:29 ` [Qemu-devel] " liu ping fan 2012-08-08 6:25 ` [PATCH 10/15] memory: change tcg related code to using PhysMap Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 6:25 ` [PATCH 11/15] lock: introduce global lock for device tree Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:41 ` Paolo Bonzini 2012-08-08 9:41 ` [Qemu-devel] " Paolo Bonzini 2012-08-09 7:28 ` liu ping fan 2012-08-09 7:28 ` [Qemu-devel] " liu ping fan 2012-08-09 7:41 ` Paolo Bonzini 2012-08-09 7:41 ` [Qemu-devel] " Paolo Bonzini 2012-08-08 9:42 ` Avi Kivity 2012-08-08 9:42 ` [Qemu-devel] " Avi Kivity 2012-08-09 7:27 ` liu ping fan 2012-08-09 7:27 ` [Qemu-devel] " liu ping fan 2012-08-09 8:31 ` Avi Kivity 2012-08-09 8:31 ` [Qemu-devel] " Avi Kivity 2012-08-08 6:25 ` [PATCH 12/15] qdev: using devtree lock to protect device's accessing Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:33 ` Peter Maydell 2012-08-08 9:33 ` [Qemu-devel] " Peter Maydell 2012-08-08 6:25 ` [PATCH 13/15] hotplug: introduce qdev_unplug_complete() to remove device from views Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:52 ` Paolo Bonzini 2012-08-08 9:52 ` [Qemu-devel] " Paolo Bonzini 2012-08-08 10:07 ` Avi Kivity 2012-08-08 10:07 ` [Qemu-devel] " Avi Kivity 2012-08-09 7:28 ` liu ping fan 2012-08-09 7:28 ` [Qemu-devel] " liu ping fan 2012-08-09 8:00 ` Paolo Bonzini 2012-08-09 8:00 ` [Qemu-devel] " Paolo Bonzini 2012-08-10 6:42 ` liu ping fan 2012-08-10 6:42 ` [Qemu-devel] " liu ping fan 2012-08-13 18:53 ` Marcelo Tosatti 2012-08-13 18:53 ` [Qemu-devel] " Marcelo Tosatti 2012-08-13 18:51 ` Marcelo Tosatti 2012-08-13 18:51 ` [Qemu-devel] " Marcelo Tosatti 2012-08-08 6:25 ` [PATCH 14/15] qom: object_unref call reclaimer Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:40 ` Paolo Bonzini 2012-08-08 9:40 ` [Qemu-devel] " Paolo Bonzini 2012-08-13 18:56 ` Marcelo Tosatti 2012-08-13 18:56 ` [Qemu-devel] " Marcelo Tosatti 2012-08-08 6:25 ` [PATCH 15/15] e1000: using new interface--unmap to unplug Liu Ping Fan 2012-08-08 6:25 ` [Qemu-devel] " Liu Ping Fan 2012-08-08 9:56 ` Paolo Bonzini 2012-08-08 9:56 ` [Qemu-devel] " Paolo Bonzini 2012-08-09 7:28 ` liu ping fan 2012-08-09 7:28 ` [Qemu-devel] " liu ping fan 2012-08-09 7:40 ` Paolo Bonzini 2012-08-09 7:40 ` [Qemu-devel] " Paolo Bonzini 2012-08-10 6:43 ` liu ping fan 2012-08-10 6:43 ` [Qemu-devel] " liu ping fan
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=5022342F.2070609@redhat.com \ --to=avi@redhat.com \ --cc=afaerber@suse.de \ --cc=anthony@codemonkey.ws \ --cc=blauwirbel@gmail.com \ --cc=jan.kiszka@siemens.com \ --cc=kvm@vger.kernel.org \ --cc=mtosatti@redhat.com \ --cc=pbonzini@redhat.com \ --cc=qemu-devel@nongnu.org \ --cc=qemulist@gmail.com \ --cc=stefanha@gmail.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.