[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v2 2/4] exec: Atomic access to map_client_list
From: |
Paolo Bonzini |
Subject: |
Re: [Qemu-devel] [PATCH v2 2/4] exec: Atomic access to map_client_list |
Date: |
Fri, 13 Mar 2015 09:11:34 +0100 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.5.0 |
On 13/03/2015 02:38, Fam Zheng wrote:
> Change map_client_list to QSLIST which supports atomic operations.
>
> There are two access points to map_client_list. One is
> cpu_register_map_client, the other is cpu_notify_map_clients called
> after releasing the global bounce buffer in address_space_unmap. Each
> is now converted to a single atomic operation.
>
> Signed-off-by: Fam Zheng <address@hidden>
> ---
> exec.c | 27 ++++++++++++---------------
> 1 file changed, 12 insertions(+), 15 deletions(-)
>
> diff --git a/exec.c b/exec.c
> index 8d4e134..93ccd5a 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -2489,11 +2489,11 @@ static BounceBuffer bounce;
> typedef struct MapClient {
> void *opaque;
> void (*callback)(void *opaque);
> - QLIST_ENTRY(MapClient) link;
> + QSLIST_ENTRY(MapClient) link;
> } MapClient;
>
> -static QLIST_HEAD(map_client_list, MapClient) map_client_list
> - = QLIST_HEAD_INITIALIZER(map_client_list);
> +static QSLIST_HEAD(map_client_list, MapClient) map_client_list
> + = QSLIST_HEAD_INITIALIZER(map_client_list);
>
> void *cpu_register_map_client(void *opaque, void (*callback)(void *opaque))
> {
> @@ -2501,26 +2501,23 @@ void *cpu_register_map_client(void *opaque, void
> (*callback)(void *opaque))
>
> client->opaque = opaque;
> client->callback = callback;
> - QLIST_INSERT_HEAD(&map_client_list, client, link);
> + QSLIST_INSERT_HEAD_ATOMIC(&map_client_list, client, link);
> return client;
> }
>
> -static void cpu_unregister_map_client(void *_client)
> -{
> - MapClient *client = (MapClient *)_client;
> -
> - QLIST_REMOVE(client, link);
> - g_free(client);
> -}
> -
> static void cpu_notify_map_clients(void)
> {
> MapClient *client;
> + QSLIST_HEAD(, MapClient) mclist;
>
> - while (!QLIST_EMPTY(&map_client_list)) {
> - client = QLIST_FIRST(&map_client_list);
> + /* Only the bounce buffer owner thread could possibly call this, but we
> + * should also coordinate with cpu_register_map_client. */
> + QSLIST_MOVE_ATOMIC(&mclist, &map_client_list);
> + while (!QSLIST_EMPTY(&mclist)) {
> + client = QSLIST_FIRST(&mclist);
> client->callback(client->opaque);
> - cpu_unregister_map_client(client);
> + QSLIST_REMOVE_HEAD(&mclist, link);
> + g_free(client);
> }
> }
>
>
Looks good.
Paolo
- Re: [Qemu-devel] [PATCH v2 4/4] dma-helpers: Move reschedule_dma BH to blk's AioContext, (continued)
[Qemu-devel] [PATCH v2 1/4] exec: Atomic access to bounce buffer, Fam Zheng, 2015/03/12
[Qemu-devel] [PATCH v2 2/4] exec: Atomic access to map_client_list, Fam Zheng, 2015/03/12
- Re: [Qemu-devel] [PATCH v2 2/4] exec: Atomic access to map_client_list,
Paolo Bonzini <=
[Qemu-devel] [PATCH v2 3/4] exec: Notify cpu_register_map_client caller if the bounce buffer is available, Fam Zheng, 2015/03/12