From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:55761) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1edkHf-0006el-NF for qemu-devel@nongnu.org; Mon, 22 Jan 2018 17:09:44 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1edkHe-0003oN-Eb for qemu-devel@nongnu.org; Mon, 22 Jan 2018 17:09:43 -0500 From: Max Reitz Date: Mon, 22 Jan 2018 23:07:59 +0100 Message-Id: <20180122220806.22154-10-mreitz@redhat.com> In-Reply-To: <20180122220806.22154-1-mreitz@redhat.com> References: <20180122220806.22154-1-mreitz@redhat.com> Subject: [Qemu-devel] [PATCH v2 09/16] block: Generalize should_update_child() rule List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-block@nongnu.org Cc: qemu-devel@nongnu.org, Max Reitz , Kevin Wolf , John Snow , Fam Zheng , Stefan Hajnoczi Currently, bdrv_replace_node() refuses to create loops from one BDS to itself if the BDS to be replaced is the backing node of the BDS to replace it: Say there is a node A and a node B. Replacing B by A means making all references to B point to A. If B is a child of A (i.e. A has a reference to B), that would mean we would have to make this reference point to A itself -- so we'd create a loop. bdrv_replace_node() (through should_update_child()) refuses to do so if B is the backing node of A. There is no reason why we should create loops if B is not the backing node of A, though. The BDS graph should never contain loops, so we should always refuse to create them. If B is a child of A and B is to be replaced by A, we should simply leave B in place there because it is the most sensible choice. A more specific argument would be: Putting filter drivers into the BDS graph is basically the same as appending an overlay to a backing chain. But the main child BDS of a filter driver is not "backing" but "file", so restricting the no-loop rule to backing nodes would fail here. Signed-off-by: Max Reitz --- include/block/block_int.h | 2 ++ block.c | 44 ++++++++++++++++++++++++++++++++++---------- 2 files changed, 36 insertions(+), 10 deletions(-) diff --git a/include/block/block_int.h b/include/block/block_int.h index 29cafa4236..03f3fdd129 100644 --- a/include/block/block_int.h +++ b/include/block/block_int.h @@ -587,6 +587,8 @@ struct BdrvChild { QLIST_ENTRY(BdrvChild) next_parent; }; +typedef QLIST_HEAD(BdrvChildList, BdrvChild) BdrvChildList; + /* * Note: the function bdrv_append() copies and swaps contents of * BlockDriverStates, so if you add new fields to this struct, please diff --git a/block.c b/block.c index a8da4f2b25..df50825d94 100644 --- a/block.c +++ b/block.c @@ -3320,16 +3320,39 @@ static bool should_update_child(BdrvChild *c, BlockDriverState *to) return false; } - if (c->role == &child_backing) { - /* If @from is a backing file of @to, ignore the child to avoid - * creating a loop. We only want to change the pointer of other - * parents. */ - QLIST_FOREACH(to_c, &to->children, next) { - if (to_c == c) { - break; - } - } - if (to_c) { + /* If the child @c belongs to the BDS @to, replacing the current + * c->bs by @to would mean to create a loop. + * + * Such a case occurs when appending a BDS to a backing chain. + * For instance, imagine the following chain: + * + * guest device -> node A -> further backing chain... + * + * Now we create a new BDS B which we want to put on top of this + * chain, so we first attach A as its backing node: + * + * node B + * | + * v + * guest device -> node A -> further backing chain... + * + * Finally we want to replace A by B. When doing that, we want to + * replace all pointers to A by pointers to B -- except for the + * pointer from B because (1) that would create a loop, and (2) + * that pointer should simply stay intact: + * + * guest device -> node B + * | + * v + * node A -> further backing chain... + * + * In general, when replacing a node A (c->bs) by a node B (@to), + * if A is a child of B, that means we cannot replace A by B there + * because that would create a loop. Silently detaching A from B + * is also not really an option. So overall just leaving A in + * place there is the most sensible choice. */ + QLIST_FOREACH(to_c, &to->children, next) { + if (to_c == c) { return false; } } @@ -3355,6 +3378,7 @@ void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to, /* Put all parents into @list and calculate their cumulative permissions */ QLIST_FOREACH_SAFE(c, &from->parents, next_parent, next) { + assert(c->bs == from); if (!should_update_child(c, to)) { continue; } -- 2.14.3