[pve-devel] [RFC PATCH kvm] add the zeroinit block driver filter

Wolfgang Bumiller w.bumiller at proxmox.com
Mon Mar 21 11:48:29 CET 2016


> On March 21, 2016 at 11:38 AM Alexandre DERUMIER <aderumier at odiso.com> wrote:
> 
> 
> >>The 'RBD' storage plugin reports that 'sparseinit' is enabled
> >>on rbd, so yes, this should apply to ceph. So unless whoever wrote
> >>that plugin made a mistake this should work for ceph, too.
> >>It'll definitely be *used* with it ;-)
> 
> Great :)
> 
> Do have plan to try to push this patch upstream to qemu ? 

Can ask, but I really doubt they'd want it seeing how it's quite a hack.
Also dangerous. If you use it on a storage that isn't zero-initialized
you'll end up with a lot of garbage ;-)

> ----- Mail original -----
> De: "Wolfgang Bumiller" <w.bumiller at proxmox.com>
> À: "aderumier" <aderumier at odiso.com>, "pve-devel" <pve-devel at pve.proxmox.com>
> Envoyé: Lundi 21 Mars 2016 11:01:47
> Objet: Re: [pve-devel] [RFC PATCH kvm] add the zeroinit block driver filter
> 
> The 'RBD' storage plugin reports that 'sparseinit' is enabled 
> on rbd, so yes, this should apply to ceph. So unless whoever wrote 
> that plugin made a mistake this should work for ceph, too. 
> It'll definitely be *used* with it ;-) 
> 
> > On March 21, 2016 at 10:21 AM Alexandre DERUMIER <aderumier at odiso.com> wrote: 
> > 
> > 
> > Hi, 
> > 
> > Could this works with ceph rbd storage as target ? 
> > 
> > currently, rbd driver don't have write zeroes implemented. 
> > 
> > 
> > ----- Mail original ----- 
> > De: "Wolfgang Bumiller" <w.bumiller at proxmox.com> 
> > À: "pve-devel" <pve-devel at pve.proxmox.com> 
> > Envoyé: Vendredi 18 Mars 2016 12:17:39 
> > Objet: [pve-devel] [RFC PATCH kvm] add the zeroinit block driver filter 
> > 
> > --- 
> > This is a block driver filter (or hack...) to force the assumption of 
> > a zero-initialized storage (has_zero_init) and skip appending 
> > write_zero() requests. 
> > 
> > We can use this for clone_disk() when converting between lvm-thin and 
> > other storage types. 
> > 
> > To use this the destination of a qemu-img or 'qmp drive-mirror' 
> > command is prefixed with 'zeroinit:' 
> > 
> > debian/patches/pve/0044-block-add-zeroinit.patch | 248 +++++++++++++++++++++++ 
> > debian/patches/series | 1 + 
> > 2 files changed, 249 insertions(+) 
> > create mode 100644 debian/patches/pve/0044-block-add-zeroinit.patch 
> > 
> > diff --git a/debian/patches/pve/0044-block-add-zeroinit.patch b/debian/patches/pve/0044-block-add-zeroinit.patch 
> > new file mode 100644 
> > index 0000000..03fc92a 
> > --- /dev/null 
> > +++ b/debian/patches/pve/0044-block-add-zeroinit.patch 
> > @@ -0,0 +1,248 @@ 
> > +From 8b996ca2740d40027652ea63f52bed8f9271331b Mon Sep 17 00:00:00 2001 
> > +From: Wolfgang Bumiller <w.bumiller at proxmox.com> 
> > +Date: Thu, 17 Mar 2016 11:33:37 +0100 
> > +Subject: [PATCH] block: add the zeroinit block driver filter 
> > + 
> > +--- 
> > + block/Makefile.objs | 1 + 
> > + block/zeroinit.c | 216 ++++++++++++++++++++++++++++++++++++++++++++++++++++ 
> > + 2 files changed, 217 insertions(+) 
> > + create mode 100644 block/zeroinit.c 
> > + 
> > +diff --git a/block/Makefile.objs b/block/Makefile.objs 
> > +index 58ef2ef..5bb55b0 100644 
> > +--- a/block/Makefile.objs 
> > ++++ b/block/Makefile.objs 
> > +@@ -4,6 +4,7 @@ block-obj-y += qed.o qed-gencb.o qed-l2-cache.o qed-table.o qed-cluster.o 
> > + block-obj-y += qed-check.o 
> > + block-obj-$(CONFIG_VHDX) += vhdx.o vhdx-endian.o vhdx-log.o 
> > + block-obj-y += quorum.o 
> > ++block-obj-y += zeroinit.o 
> > + block-obj-y += parallels.o blkdebug.o blkverify.o 
> > + block-obj-y += block-backend.o snapshot.o qapi.o 
> > + block-obj-$(CONFIG_WIN32) += raw-win32.o win32-aio.o 
> > +diff --git a/block/zeroinit.c b/block/zeroinit.c 
> > +new file mode 100644 
> > +index 0000000..04a9bd0 
> > +--- /dev/null 
> > ++++ b/block/zeroinit.c 
> > +@@ -0,0 +1,216 @@ 
> > ++/* 
> > ++ * Filter to fake a zero-initialized block device. 
> > ++ * 
> > ++ * Copyright (c) 2016 Wolfgang Bumiller <w.bumiller at proxmox.com> 
> > ++ * Copyright (c) 2016 Proxmox Server Solutions GmbH 
> > ++ * 
> > ++ * This work is licensed under the terms of the GNU GPL, version 2 or later. 
> > ++ * See the COPYING file in the top-level directory. 
> > ++ */ 
> > ++ 
> > ++#include "block/block_int.h" 
> > ++#include "qapi/qmp/qdict.h" 
> > ++#include "qapi/qmp/qstring.h" 
> > ++ 
> > ++typedef struct { 
> > ++ bool has_zero_init; 
> > ++ int64_t extents; 
> > ++} BDRVZeroinitState; 
> > ++ 
> > ++/* Valid blkverify filenames look like blkverify:path/to/raw_image:path/to/image */ 
> > ++static void zeroinit_parse_filename(const char *filename, QDict *options, 
> > ++ Error **errp) 
> > ++{ 
> > ++ QString *raw_path; 
> > ++ 
> > ++ /* Parse the blkverify: prefix */ 
> > ++ if (!strstart(filename, "zeroinit:", &filename)) { 
> > ++ /* There was no prefix; therefore, all options have to be already 
> > ++ present in the QDict (except for the filename) */ 
> > ++ return; 
> > ++ } 
> > ++ 
> > ++ raw_path = qstring_from_str(filename); 
> > ++ qdict_put(options, "x-next", raw_path); 
> > ++} 
> > ++ 
> > ++static QemuOptsList runtime_opts = { 
> > ++ .name = "zeroinit", 
> > ++ .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head), 
> > ++ .desc = { 
> > ++ { 
> > ++ .name = "x-next", 
> > ++ .type = QEMU_OPT_STRING, 
> > ++ .help = "[internal use only, will be removed]", 
> > ++ }, 
> > ++ { 
> > ++ .name = "x-zeroinit", 
> > ++ .type = QEMU_OPT_BOOL, 
> > ++ .help = "set has_initialized_zero flag", 
> > ++ }, 
> > ++ { /* end of list */ } 
> > ++ }, 
> > ++}; 
> > ++ 
> > ++static int zeroinit_open(BlockDriverState *bs, QDict *options, int flags, 
> > ++ Error **errp) 
> > ++{ 
> > ++ BDRVZeroinitState *s = bs->opaque; 
> > ++ QemuOpts *opts; 
> > ++ Error *local_err = NULL; 
> > ++ int ret; 
> > ++ 
> > ++ s->extents = 0; 
> > ++ 
> > ++ opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort); 
> > ++ qemu_opts_absorb_qdict(opts, options, &local_err); 
> > ++ if (local_err) { 
> > ++ error_propagate(errp, local_err); 
> > ++ ret = -EINVAL; 
> > ++ goto fail; 
> > ++ } 
> > ++ 
> > ++ /* Open the raw file */ 
> > ++ bs->file = bdrv_open_child(qemu_opt_get(opts, "x-next"), options, "next", 
> > ++ bs, &child_file, false, &local_err); 
> > ++ if (local_err) { 
> > ++ ret = -EINVAL; 
> > ++ error_propagate(errp, local_err); 
> > ++ goto fail; 
> > ++ } 
> > ++ 
> > ++ /* set the options */ 
> > ++ s->has_zero_init = qemu_opt_get_bool(opts, "x-zeroinit", true); 
> > ++ 
> > ++ ret = 0; 
> > ++fail: 
> > ++ if (ret < 0) { 
> > ++ bdrv_unref_child(bs, bs->file); 
> > ++ } 
> > ++ qemu_opts_del(opts); 
> > ++ return ret; 
> > ++} 
> > ++ 
> > ++static void zeroinit_close(BlockDriverState *bs) 
> > ++{ 
> > ++ BDRVZeroinitState *s = bs->opaque; 
> > ++ (void)s; 
> > ++} 
> > ++ 
> > ++static int64_t zeroinit_getlength(BlockDriverState *bs) 
> > ++{ 
> > ++ return bdrv_getlength(bs->file->bs); 
> > ++} 
> > ++ 
> > ++static BlockAIOCB *zeroinit_aio_readv(BlockDriverState *bs, 
> > ++ int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, 
> > ++ BlockCompletionFunc *cb, void *opaque) 
> > ++{ 
> > ++ return bdrv_aio_readv(bs->file->bs, sector_num, qiov, nb_sectors, 
> > ++ cb, opaque); 
> > ++} 
> > ++ 
> > ++static int coroutine_fn zeroinit_co_write_zeroes(BlockDriverState *bs, int64_t sector_num, 
> > ++ int nb_sectors, BdrvRequestFlags flags) 
> > ++{ 
> > ++ BDRVZeroinitState *s = bs->opaque; 
> > ++ if (sector_num >= s->extents) 
> > ++ return 0; 
> > ++ return bdrv_write_zeroes(bs->file->bs, sector_num, nb_sectors, flags); 
> > ++} 
> > ++ 
> > ++static BlockAIOCB *zeroinit_aio_writev(BlockDriverState *bs, 
> > ++ int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, 
> > ++ BlockCompletionFunc *cb, void *opaque) 
> > ++{ 
> > ++ BDRVZeroinitState *s = bs->opaque; 
> > ++ int64_t extents = sector_num + nb_sectors + 1; 
> > ++ if (extents > s->extents) 
> > ++ s->extents = extents; 
> > ++ return bdrv_aio_writev(bs->file->bs, sector_num, qiov, nb_sectors, 
> > ++ cb, opaque); 
> > ++} 
> > ++ 
> > ++static BlockAIOCB *zeroinit_aio_flush(BlockDriverState *bs, 
> > ++ BlockCompletionFunc *cb, 
> > ++ void *opaque) 
> > ++{ 
> > ++ return bdrv_aio_flush(bs->file->bs, cb, opaque); 
> > ++} 
> > ++ 
> > ++static bool zeroinit_recurse_is_first_non_filter(BlockDriverState *bs, 
> > ++ BlockDriverState *candidate) 
> > ++{ 
> > ++ return bdrv_recurse_is_first_non_filter(bs->file->bs, candidate); 
> > ++} 
> > ++ 
> > ++static coroutine_fn int zeroinit_co_flush(BlockDriverState *bs) 
> > ++{ 
> > ++ return bdrv_co_flush(bs->file->bs); 
> > ++} 
> > ++ 
> > ++static int zeroinit_has_zero_init(BlockDriverState *bs) 
> > ++{ 
> > ++ BDRVZeroinitState *s = bs->opaque; 
> > ++ return s->has_zero_init; 
> > ++} 
> > ++ 
> > ++static int64_t coroutine_fn zeroinit_co_get_block_status(BlockDriverState *bs, 
> > ++ int64_t sector_num, 
> > ++ int nb_sectors, int *pnum) 
> > ++{ 
> > ++ return bdrv_get_block_status(bs->file->bs, sector_num, nb_sectors, pnum); 
> > ++} 
> > ++ 
> > ++static coroutine_fn BlockAIOCB *zeroinit_aio_discard(BlockDriverState *bs, 
> > ++ int64_t sector_num, int nb_sectors, 
> > ++ BlockCompletionFunc *cb, void *opaque) 
> > ++{ 
> > ++ return bdrv_aio_discard(bs->file->bs, sector_num, nb_sectors, cb, opaque); 
> > ++} 
> > ++ 
> > ++static int zeroinit_truncate(BlockDriverState *bs, int64_t offset) 
> > ++{ 
> > ++ return bdrv_truncate(bs->file->bs, offset); 
> > ++} 
> > ++ 
> > ++static int zeroinit_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) 
> > ++{ 
> > ++ return bdrv_get_info(bs->file->bs, bdi); 
> > ++} 
> > ++ 
> > ++static BlockDriver bdrv_zeroinit = { 
> > ++ .format_name = "zeroinit", 
> > ++ .protocol_name = "zeroinit", 
> > ++ .instance_size = sizeof(BDRVZeroinitState), 
> > ++ 
> > ++ .bdrv_parse_filename = zeroinit_parse_filename, 
> > ++ .bdrv_file_open = zeroinit_open, 
> > ++ .bdrv_close = zeroinit_close, 
> > ++ .bdrv_getlength = zeroinit_getlength, 
> > ++ .bdrv_co_flush_to_disk = zeroinit_co_flush, 
> > ++ 
> > ++ .bdrv_co_write_zeroes = zeroinit_co_write_zeroes, 
> > ++ .bdrv_aio_writev = zeroinit_aio_writev, 
> > ++ .bdrv_aio_readv = zeroinit_aio_readv, 
> > ++ .bdrv_aio_flush = zeroinit_aio_flush, 
> > ++ 
> > ++ .is_filter = true, 
> > ++ .bdrv_recurse_is_first_non_filter = zeroinit_recurse_is_first_non_filter, 
> > ++ 
> > ++ .bdrv_has_zero_init = zeroinit_has_zero_init, 
> > ++ 
> > ++ .bdrv_co_get_block_status = zeroinit_co_get_block_status, 
> > ++ 
> > ++ .bdrv_aio_discard = zeroinit_aio_discard, 
> > ++ 
> > ++ .bdrv_truncate = zeroinit_truncate, 
> > ++ .bdrv_get_info = zeroinit_get_info, 
> > ++}; 
> > ++ 
> > ++static void bdrv_zeroinit_init(void) 
> > ++{ 
> > ++ bdrv_register(&bdrv_zeroinit); 
> > ++} 
> > ++ 
> > ++block_init(bdrv_zeroinit_init); 
> > +-- 
> > +2.1.4 
> > + 
> > diff --git a/debian/patches/series b/debian/patches/series 
> > index f07e028..e8f1710 100644 
> > --- a/debian/patches/series 
> > +++ b/debian/patches/series 
> > @@ -41,6 +41,7 @@ pve/0040-vnc-make-x509-imply-tls-again.patch 
> > pve/0041-PVE-VNC-authentication.patch 
> > pve/0042-vma-writer-don-t-bail-out-on-zero-length-files.patch 
> > pve/0043-vma-better-driver-guessing-for-bdrv_open.patch 
> > +pve/0044-block-add-zeroinit.patch 
> > extra/CVE-2015-8558-ehci_make_idt_processing_more_robust.patch 
> > extra/CVE-2015-8613-scsi-initialize-info-object.patch 
> > extra/CVE-2015-8701-net-rocker-off-by-one.patch 
> > -- 
> > 2.1.4 
> > 
> > 
> > _______________________________________________ 
> > pve-devel mailing list 
> > pve-devel at pve.proxmox.com 
> > http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
> > 
> > _______________________________________________ 
> > pve-devel mailing list 
> > pve-devel at pve.proxmox.com 
> > http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
>




More information about the pve-devel mailing list