* [PULL v2 0/8] 9p patches 2019-10-10
@ 2019-10-10 13:18 Greg Kurz
2019-10-10 13:18 ` [PULL v2 1/8] 9p: unsigned type for type, version, path Greg Kurz
` (8 more replies)
0 siblings, 9 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel; +Cc: Peter Maydell, Greg Kurz
The following changes since commit 98b2e3c9ab3abfe476a2b02f8f51813edb90e72d:
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging (2019-10-08 16:08:35 +0100)
are available in the Git repository at:
https://github.com/gkurz/qemu.git tags/9p-next-2019-10-10
for you to fetch changes up to e410bbc52a7a246858af54ce1842fadf26f39ff7:
MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes" (2019-10-10 12:36:28 +0200)
----------------------------------------------------------------
The most notable change is that we now detect cross-device setups in the
host since it may cause inode number collision and mayhem in the guest.
A new fsdev property is added for the user to choose the appropriate
policy to handle that: either remap all inode numbers or fail I/Os to
another host device or just print out a warning (default behaviour).
This is also my last PR as _active_ maintainer of 9pfs.
----------------------------------------------------------------
Antonios Motakis (4):
9p: unsigned type for type, version, path
9p: Treat multiple devices on one export as an error
9p: Added virtfs option 'multidevs=remap|forbid|warn'
9p: stat_to_qid: implement slow path
Christian Schoenebeck (1):
9p: Use variable length suffixes for inode remapping
Greg Kurz (3):
9p: Simplify error path of v9fs_device_realize_common()
fsdev: Add return value to fsdev_throttle_parse_opts()
MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes"
MAINTAINERS | 2 +-
fsdev/9p-marshal.h | 6 +-
fsdev/file-op-9p.h | 5 +
fsdev/qemu-fsdev-opts.c | 7 +-
fsdev/qemu-fsdev-throttle.c | 4 +-
fsdev/qemu-fsdev-throttle.h | 2 +-
fsdev/qemu-fsdev.c | 1 +
hw/9pfs/9p-local.c | 28 ++-
hw/9pfs/9p-proxy.c | 4 +
hw/9pfs/9p.c | 534 ++++++++++++++++++++++++++++++++++++++++----
hw/9pfs/9p.h | 60 +++++
hw/9pfs/trace-events | 14 +-
qemu-options.hx | 26 ++-
vl.c | 7 +-
14 files changed, 639 insertions(+), 61 deletions(-)
--
2.21.0
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PULL v2 1/8] 9p: unsigned type for type, version, path
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-10 13:18 ` [PULL v2 2/8] 9p: Simplify error path of v9fs_device_realize_common() Greg Kurz
` (7 subsequent siblings)
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel
Cc: Peter Maydell, Christian Schoenebeck, Greg Kurz, Antonios Motakis
From: Antonios Motakis <antonios.motakis@huawei.com>
There is no need for signedness on these QID fields for 9p.
Signed-off-by: Antonios Motakis <antonios.motakis@huawei.com>
[CS: - Also make QID type unsigned.
- Adjust donttouch_stat() to new types.
- Adjust trace-events to new types. ]
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
---
fsdev/9p-marshal.h | 6 +++---
hw/9pfs/9p.c | 6 +++---
hw/9pfs/trace-events | 14 +++++++-------
3 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/fsdev/9p-marshal.h b/fsdev/9p-marshal.h
index c8823d878f98..8f3babb60a05 100644
--- a/fsdev/9p-marshal.h
+++ b/fsdev/9p-marshal.h
@@ -9,9 +9,9 @@ typedef struct V9fsString
typedef struct V9fsQID
{
- int8_t type;
- int32_t version;
- int64_t path;
+ uint8_t type;
+ uint32_t version;
+ uint64_t path;
} V9fsQID;
typedef struct V9fsStat
diff --git a/hw/9pfs/9p.c b/hw/9pfs/9p.c
index cce2366219be..cf317bdd2b92 100644
--- a/hw/9pfs/9p.c
+++ b/hw/9pfs/9p.c
@@ -744,9 +744,9 @@ static int donttouch_stat(V9fsStat *stat)
{
if (stat->type == -1 &&
stat->dev == -1 &&
- stat->qid.type == -1 &&
- stat->qid.version == -1 &&
- stat->qid.path == -1 &&
+ stat->qid.type == 0xff &&
+ stat->qid.version == (uint32_t) -1 &&
+ stat->qid.path == (uint64_t) -1 &&
stat->mode == -1 &&
stat->atime == -1 &&
stat->mtime == -1 &&
diff --git a/hw/9pfs/trace-events b/hw/9pfs/trace-events
index c0a0a4ab5de8..10188daf7fa5 100644
--- a/hw/9pfs/trace-events
+++ b/hw/9pfs/trace-events
@@ -6,7 +6,7 @@ v9fs_rerror(uint16_t tag, uint8_t id, int err) "tag %d id %d err %d"
v9fs_version(uint16_t tag, uint8_t id, int32_t msize, char* version) "tag %d id %d msize %d version %s"
v9fs_version_return(uint16_t tag, uint8_t id, int32_t msize, char* version) "tag %d id %d msize %d version %s"
v9fs_attach(uint16_t tag, uint8_t id, int32_t fid, int32_t afid, char* uname, char* aname) "tag %u id %u fid %d afid %d uname %s aname %s"
-v9fs_attach_return(uint16_t tag, uint8_t id, int8_t type, int32_t version, int64_t path) "tag %d id %d type %d version %d path %"PRId64
+v9fs_attach_return(uint16_t tag, uint8_t id, uint8_t type, uint32_t version, uint64_t path) "tag %u id %u type %u version %u path %"PRIu64
v9fs_stat(uint16_t tag, uint8_t id, int32_t fid) "tag %d id %d fid %d"
v9fs_stat_return(uint16_t tag, uint8_t id, int32_t mode, int32_t atime, int32_t mtime, int64_t length) "tag %d id %d stat={mode %d atime %d mtime %d length %"PRId64"}"
v9fs_getattr(uint16_t tag, uint8_t id, int32_t fid, uint64_t request_mask) "tag %d id %d fid %d request_mask %"PRIu64
@@ -14,9 +14,9 @@ v9fs_getattr_return(uint16_t tag, uint8_t id, uint64_t result_mask, uint32_t mod
v9fs_walk(uint16_t tag, uint8_t id, int32_t fid, int32_t newfid, uint16_t nwnames) "tag %d id %d fid %d newfid %d nwnames %d"
v9fs_walk_return(uint16_t tag, uint8_t id, uint16_t nwnames, void* qids) "tag %d id %d nwnames %d qids %p"
v9fs_open(uint16_t tag, uint8_t id, int32_t fid, int32_t mode) "tag %d id %d fid %d mode %d"
-v9fs_open_return(uint16_t tag, uint8_t id, int8_t type, int32_t version, int64_t path, int iounit) "tag %d id %d qid={type %d version %d path %"PRId64"} iounit %d"
+v9fs_open_return(uint16_t tag, uint8_t id, uint8_t type, uint32_t version, uint64_t path, int iounit) "tag %u id %u qid={type %u version %u path %"PRIu64"} iounit %d"
v9fs_lcreate(uint16_t tag, uint8_t id, int32_t dfid, int32_t flags, int32_t mode, uint32_t gid) "tag %d id %d dfid %d flags %d mode %d gid %u"
-v9fs_lcreate_return(uint16_t tag, uint8_t id, int8_t type, int32_t version, int64_t path, int32_t iounit) "tag %d id %d qid={type %d version %d path %"PRId64"} iounit %d"
+v9fs_lcreate_return(uint16_t tag, uint8_t id, uint8_t type, uint32_t version, uint64_t path, int32_t iounit) "tag %u id %u qid={type %u version %u path %"PRIu64"} iounit %d"
v9fs_fsync(uint16_t tag, uint8_t id, int32_t fid, int datasync) "tag %d id %d fid %d datasync %d"
v9fs_clunk(uint16_t tag, uint8_t id, int32_t fid) "tag %d id %d fid %d"
v9fs_read(uint16_t tag, uint8_t id, int32_t fid, uint64_t off, uint32_t max_count) "tag %d id %d fid %d off %"PRIu64" max_count %u"
@@ -26,21 +26,21 @@ v9fs_readdir_return(uint16_t tag, uint8_t id, uint32_t count, ssize_t retval) "t
v9fs_write(uint16_t tag, uint8_t id, int32_t fid, uint64_t off, uint32_t count, int cnt) "tag %d id %d fid %d off %"PRIu64" count %u cnt %d"
v9fs_write_return(uint16_t tag, uint8_t id, int32_t total, ssize_t err) "tag %d id %d total %d err %zd"
v9fs_create(uint16_t tag, uint8_t id, int32_t fid, char* name, int32_t perm, int8_t mode) "tag %d id %d fid %d name %s perm %d mode %d"
-v9fs_create_return(uint16_t tag, uint8_t id, int8_t type, int32_t version, int64_t path, int iounit) "tag %d id %d qid={type %d version %d path %"PRId64"} iounit %d"
+v9fs_create_return(uint16_t tag, uint8_t id, uint8_t type, uint32_t version, uint64_t path, int iounit) "tag %u id %u qid={type %u version %u path %"PRIu64"} iounit %d"
v9fs_symlink(uint16_t tag, uint8_t id, int32_t fid, char* name, char* symname, uint32_t gid) "tag %d id %d fid %d name %s symname %s gid %u"
-v9fs_symlink_return(uint16_t tag, uint8_t id, int8_t type, int32_t version, int64_t path) "tag %d id %d qid={type %d version %d path %"PRId64"}"
+v9fs_symlink_return(uint16_t tag, uint8_t id, uint8_t type, uint32_t version, uint64_t path) "tag %u id %u qid={type %u version %u path %"PRIu64"}"
v9fs_flush(uint16_t tag, uint8_t id, int16_t flush_tag) "tag %d id %d flush_tag %d"
v9fs_link(uint16_t tag, uint8_t id, int32_t dfid, int32_t oldfid, char* name) "tag %d id %d dfid %d oldfid %d name %s"
v9fs_remove(uint16_t tag, uint8_t id, int32_t fid) "tag %d id %d fid %d"
v9fs_wstat(uint16_t tag, uint8_t id, int32_t fid, int32_t mode, int32_t atime, int32_t mtime) "tag %u id %u fid %d stat={mode %d atime %d mtime %d}"
v9fs_mknod(uint16_t tag, uint8_t id, int32_t fid, int mode, int major, int minor) "tag %d id %d fid %d mode %d major %d minor %d"
-v9fs_mknod_return(uint16_t tag, uint8_t id, int8_t type, int32_t version, int64_t path) "tag %d id %d qid={type %d version %d path %"PRId64"}"
+v9fs_mknod_return(uint16_t tag, uint8_t id, uint8_t type, uint32_t version, uint64_t path) "tag %u id %u qid={type %u version %u path %"PRIu64"}"
v9fs_lock(uint16_t tag, uint8_t id, int32_t fid, uint8_t type, uint64_t start, uint64_t length) "tag %d id %d fid %d type %d start %"PRIu64" length %"PRIu64
v9fs_lock_return(uint16_t tag, uint8_t id, int8_t status) "tag %d id %d status %d"
v9fs_getlock(uint16_t tag, uint8_t id, int32_t fid, uint8_t type, uint64_t start, uint64_t length)"tag %d id %d fid %d type %d start %"PRIu64" length %"PRIu64
v9fs_getlock_return(uint16_t tag, uint8_t id, uint8_t type, uint64_t start, uint64_t length, uint32_t proc_id) "tag %d id %d type %d start %"PRIu64" length %"PRIu64" proc_id %u"
v9fs_mkdir(uint16_t tag, uint8_t id, int32_t fid, char* name, int mode, uint32_t gid) "tag %u id %u fid %d name %s mode %d gid %u"
-v9fs_mkdir_return(uint16_t tag, uint8_t id, int8_t type, int32_t version, int64_t path, int err) "tag %u id %u qid={type %d version %d path %"PRId64"} err %d"
+v9fs_mkdir_return(uint16_t tag, uint8_t id, uint8_t type, uint32_t version, uint64_t path, int err) "tag %u id %u qid={type %u version %u path %"PRIu64"} err %d"
v9fs_xattrwalk(uint16_t tag, uint8_t id, int32_t fid, int32_t newfid, char* name) "tag %d id %d fid %d newfid %d name %s"
v9fs_xattrwalk_return(uint16_t tag, uint8_t id, int64_t size) "tag %d id %d size %"PRId64
v9fs_xattrcreate(uint16_t tag, uint8_t id, int32_t fid, char* name, uint64_t size, int flags) "tag %d id %d fid %d name %s size %"PRIu64" flags %d"
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PULL v2 2/8] 9p: Simplify error path of v9fs_device_realize_common()
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
2019-10-10 13:18 ` [PULL v2 1/8] 9p: unsigned type for type, version, path Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-10 13:18 ` [PULL v2 3/8] fsdev: Add return value to fsdev_throttle_parse_opts() Greg Kurz
` (6 subsequent siblings)
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel; +Cc: Peter Maydell, Greg Kurz
Make v9fs_device_unrealize_common() idempotent and use it for rollback,
in order to reduce code duplication.
Signed-off-by: Greg Kurz <groug@kaod.org>
---
hw/9pfs/9p-local.c | 4 ++++
hw/9pfs/9p-proxy.c | 4 ++++
hw/9pfs/9p.c | 16 ++++++----------
3 files changed, 14 insertions(+), 10 deletions(-)
diff --git a/hw/9pfs/9p-local.c b/hw/9pfs/9p-local.c
index 08e673a79c22..6f7309f4e691 100644
--- a/hw/9pfs/9p-local.c
+++ b/hw/9pfs/9p-local.c
@@ -1465,6 +1465,10 @@ static void local_cleanup(FsContext *ctx)
{
LocalData *data = ctx->private;
+ if (!data) {
+ return;
+ }
+
close(data->mountfd);
g_free(data);
}
diff --git a/hw/9pfs/9p-proxy.c b/hw/9pfs/9p-proxy.c
index 57a8c1c80886..97ab9c58a573 100644
--- a/hw/9pfs/9p-proxy.c
+++ b/hw/9pfs/9p-proxy.c
@@ -1185,6 +1185,10 @@ static void proxy_cleanup(FsContext *ctx)
{
V9fsProxy *proxy = ctx->private;
+ if (!proxy) {
+ return;
+ }
+
g_free(proxy->out_iovec.iov_base);
g_free(proxy->in_iovec.iov_base);
if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
diff --git a/hw/9pfs/9p.c b/hw/9pfs/9p.c
index cf317bdd2b92..ba1ab920f1eb 100644
--- a/hw/9pfs/9p.c
+++ b/hw/9pfs/9p.c
@@ -3637,27 +3637,23 @@ int v9fs_device_realize_common(V9fsState *s, const V9fsTransport *t,
s->ctx.fst = &fse->fst;
fsdev_throttle_init(s->ctx.fst);
- v9fs_path_free(&path);
-
rc = 0;
out:
if (rc) {
- if (s->ops && s->ops->cleanup && s->ctx.private) {
- s->ops->cleanup(&s->ctx);
- }
- g_free(s->tag);
- g_free(s->ctx.fs_root);
- v9fs_path_free(&path);
+ v9fs_device_unrealize_common(s, NULL);
}
+ v9fs_path_free(&path);
return rc;
}
void v9fs_device_unrealize_common(V9fsState *s, Error **errp)
{
- if (s->ops->cleanup) {
+ if (s->ops && s->ops->cleanup) {
s->ops->cleanup(&s->ctx);
}
- fsdev_throttle_cleanup(s->ctx.fst);
+ if (s->ctx.fst) {
+ fsdev_throttle_cleanup(s->ctx.fst);
+ }
g_free(s->tag);
g_free(s->ctx.fs_root);
}
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PULL v2 3/8] fsdev: Add return value to fsdev_throttle_parse_opts()
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
2019-10-10 13:18 ` [PULL v2 1/8] 9p: unsigned type for type, version, path Greg Kurz
2019-10-10 13:18 ` [PULL v2 2/8] 9p: Simplify error path of v9fs_device_realize_common() Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-10 13:18 ` [PULL v2 4/8] 9p: Treat multiple devices on one export as an error Greg Kurz
` (5 subsequent siblings)
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel; +Cc: Peter Maydell, Greg Kurz
It is more convenient to use the return value of the function to notify
errors, rather than to be tied up setting up the &local_err boilerplate.
Signed-off-by: Greg Kurz <groug@kaod.org>
---
fsdev/qemu-fsdev-throttle.c | 4 ++--
fsdev/qemu-fsdev-throttle.h | 2 +-
hw/9pfs/9p-local.c | 3 +--
3 files changed, 4 insertions(+), 5 deletions(-)
diff --git a/fsdev/qemu-fsdev-throttle.c b/fsdev/qemu-fsdev-throttle.c
index fe1d76dc76e7..5c83a1cc09c3 100644
--- a/fsdev/qemu-fsdev-throttle.c
+++ b/fsdev/qemu-fsdev-throttle.c
@@ -31,7 +31,7 @@ static void fsdev_throttle_write_timer_cb(void *opaque)
qemu_co_enter_next(&fst->throttled_reqs[true], NULL);
}
-void fsdev_throttle_parse_opts(QemuOpts *opts, FsThrottle *fst, Error **errp)
+int fsdev_throttle_parse_opts(QemuOpts *opts, FsThrottle *fst, Error **errp)
{
throttle_config_init(&fst->cfg);
fst->cfg.buckets[THROTTLE_BPS_TOTAL].avg =
@@ -75,7 +75,7 @@ void fsdev_throttle_parse_opts(QemuOpts *opts, FsThrottle *fst, Error **errp)
fst->cfg.op_size =
qemu_opt_get_number(opts, "throttling.iops-size", 0);
- throttle_is_valid(&fst->cfg, errp);
+ return throttle_is_valid(&fst->cfg, errp) ? 0 : -1;
}
void fsdev_throttle_init(FsThrottle *fst)
diff --git a/fsdev/qemu-fsdev-throttle.h b/fsdev/qemu-fsdev-throttle.h
index c98e2feab553..a21aecddc734 100644
--- a/fsdev/qemu-fsdev-throttle.h
+++ b/fsdev/qemu-fsdev-throttle.h
@@ -26,7 +26,7 @@ typedef struct FsThrottle {
CoQueue throttled_reqs[2];
} FsThrottle;
-void fsdev_throttle_parse_opts(QemuOpts *, FsThrottle *, Error **);
+int fsdev_throttle_parse_opts(QemuOpts *, FsThrottle *, Error **);
void fsdev_throttle_init(FsThrottle *);
diff --git a/hw/9pfs/9p-local.c b/hw/9pfs/9p-local.c
index 6f7309f4e691..5c7f4cd9c7b0 100644
--- a/hw/9pfs/9p-local.c
+++ b/hw/9pfs/9p-local.c
@@ -1511,8 +1511,7 @@ static int local_parse_opts(QemuOpts *opts, FsDriverEntry *fse, Error **errp)
return -1;
}
- fsdev_throttle_parse_opts(opts, &fse->fst, &local_err);
- if (local_err) {
+ if (fsdev_throttle_parse_opts(opts, &fse->fst, &local_err)) {
error_propagate_prepend(errp, local_err,
"invalid throttle configuration: ");
return -1;
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PULL v2 4/8] 9p: Treat multiple devices on one export as an error
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
` (2 preceding siblings ...)
2019-10-10 13:18 ` [PULL v2 3/8] fsdev: Add return value to fsdev_throttle_parse_opts() Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-10 13:18 ` [PULL v2 5/8] 9p: Added virtfs option 'multidevs=remap|forbid|warn' Greg Kurz
` (4 subsequent siblings)
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel
Cc: Peter Maydell, Christian Schoenebeck, Greg Kurz, Antonios Motakis
From: Antonios Motakis <antonios.motakis@huawei.com>
The QID path should uniquely identify a file. However, the
inode of a file is currently used as the QID path, which
on its own only uniquely identifies files within a device.
Here we track the device hosting the 9pfs share, in order
to prevent security issues with QID path collisions from
other devices.
We only print a warning for now but a subsequent patch will
allow users to have finer control over the desired behaviour.
Failing the I/O will be one the proposed behaviour, so we
also change stat_to_qid() to return an error here in order to
keep other patches simpler.
Signed-off-by: Antonios Motakis <antonios.motakis@huawei.com>
[CS: - Assign dev_id to export root's device already in
v9fs_device_realize_common(), not postponed in
stat_to_qid().
- error_report_once() if more than one device was
shared by export.
- Return -ENODEV instead of -ENOSYS in stat_to_qid().
- Fixed typo in log comment. ]
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
[groug, changed to warning, updated message and changelog]
Signed-off-by: Greg Kurz <groug@kaod.org>
---
hw/9pfs/9p.c | 70 +++++++++++++++++++++++++++++++++++++++++-----------
hw/9pfs/9p.h | 1 +
2 files changed, 57 insertions(+), 14 deletions(-)
diff --git a/hw/9pfs/9p.c b/hw/9pfs/9p.c
index ba1ab920f1eb..5a895ae0bbfe 100644
--- a/hw/9pfs/9p.c
+++ b/hw/9pfs/9p.c
@@ -573,10 +573,19 @@ static void coroutine_fn virtfs_reset(V9fsPDU *pdu)
P9_STAT_MODE_SOCKET)
/* This is the algorithm from ufs in spfs */
-static void stat_to_qid(const struct stat *stbuf, V9fsQID *qidp)
+static int stat_to_qid(V9fsPDU *pdu, const struct stat *stbuf, V9fsQID *qidp)
{
size_t size;
+ if (pdu->s->dev_id != stbuf->st_dev) {
+ warn_report_once(
+ "9p: Multiple devices detected in same VirtFS export, "
+ "which might lead to file ID collisions and severe "
+ "misbehaviours on guest! You should use a separate "
+ "export for each device shared from host."
+ );
+ }
+
memset(&qidp->path, 0, sizeof(qidp->path));
size = MIN(sizeof(stbuf->st_ino), sizeof(qidp->path));
memcpy(&qidp->path, &stbuf->st_ino, size);
@@ -588,6 +597,8 @@ static void stat_to_qid(const struct stat *stbuf, V9fsQID *qidp)
if (S_ISLNK(stbuf->st_mode)) {
qidp->type |= P9_QID_TYPE_SYMLINK;
}
+
+ return 0;
}
static int coroutine_fn fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp,
@@ -600,7 +611,10 @@ static int coroutine_fn fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp,
if (err < 0) {
return err;
}
- stat_to_qid(&stbuf, qidp);
+ err = stat_to_qid(pdu, &stbuf, qidp);
+ if (err < 0) {
+ return err;
+ }
return 0;
}
@@ -831,7 +845,10 @@ static int coroutine_fn stat_to_v9stat(V9fsPDU *pdu, V9fsPath *path,
memset(v9stat, 0, sizeof(*v9stat));
- stat_to_qid(stbuf, &v9stat->qid);
+ err = stat_to_qid(pdu, stbuf, &v9stat->qid);
+ if (err < 0) {
+ return err;
+ }
v9stat->mode = stat_to_v9mode(stbuf);
v9stat->atime = stbuf->st_atime;
v9stat->mtime = stbuf->st_mtime;
@@ -892,7 +909,7 @@ static int coroutine_fn stat_to_v9stat(V9fsPDU *pdu, V9fsPath *path,
#define P9_STATS_ALL 0x00003fffULL /* Mask for All fields above */
-static void stat_to_v9stat_dotl(V9fsState *s, const struct stat *stbuf,
+static int stat_to_v9stat_dotl(V9fsPDU *pdu, const struct stat *stbuf,
V9fsStatDotl *v9lstat)
{
memset(v9lstat, 0, sizeof(*v9lstat));
@@ -914,7 +931,7 @@ static void stat_to_v9stat_dotl(V9fsState *s, const struct stat *stbuf,
/* Currently we only support BASIC fields in stat */
v9lstat->st_result_mask = P9_STATS_BASIC;
- stat_to_qid(stbuf, &v9lstat->qid);
+ return stat_to_qid(pdu, stbuf, &v9lstat->qid);
}
static void print_sg(struct iovec *sg, int cnt)
@@ -1116,7 +1133,6 @@ static void coroutine_fn v9fs_getattr(void *opaque)
uint64_t request_mask;
V9fsStatDotl v9stat_dotl;
V9fsPDU *pdu = opaque;
- V9fsState *s = pdu->s;
retval = pdu_unmarshal(pdu, offset, "dq", &fid, &request_mask);
if (retval < 0) {
@@ -1137,7 +1153,10 @@ static void coroutine_fn v9fs_getattr(void *opaque)
if (retval < 0) {
goto out;
}
- stat_to_v9stat_dotl(s, &stbuf, &v9stat_dotl);
+ retval = stat_to_v9stat_dotl(pdu, &stbuf, &v9stat_dotl);
+ if (retval < 0) {
+ goto out;
+ }
/* fill st_gen if requested and supported by underlying fs */
if (request_mask & P9_STATS_GEN) {
@@ -1382,7 +1401,10 @@ static void coroutine_fn v9fs_walk(void *opaque)
if (err < 0) {
goto out;
}
- stat_to_qid(&stbuf, &qid);
+ err = stat_to_qid(pdu, &stbuf, &qid);
+ if (err < 0) {
+ goto out;
+ }
v9fs_path_copy(&dpath, &path);
}
memcpy(&qids[name_idx], &qid, sizeof(qid));
@@ -1484,7 +1506,10 @@ static void coroutine_fn v9fs_open(void *opaque)
if (err < 0) {
goto out;
}
- stat_to_qid(&stbuf, &qid);
+ err = stat_to_qid(pdu, &stbuf, &qid);
+ if (err < 0) {
+ goto out;
+ }
if (S_ISDIR(stbuf.st_mode)) {
err = v9fs_co_opendir(pdu, fidp);
if (err < 0) {
@@ -1594,7 +1619,10 @@ static void coroutine_fn v9fs_lcreate(void *opaque)
fidp->flags |= FID_NON_RECLAIMABLE;
}
iounit = get_iounit(pdu, &fidp->path);
- stat_to_qid(&stbuf, &qid);
+ err = stat_to_qid(pdu, &stbuf, &qid);
+ if (err < 0) {
+ goto out;
+ }
err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
if (err < 0) {
goto out;
@@ -2328,7 +2356,10 @@ static void coroutine_fn v9fs_create(void *opaque)
}
}
iounit = get_iounit(pdu, &fidp->path);
- stat_to_qid(&stbuf, &qid);
+ err = stat_to_qid(pdu, &stbuf, &qid);
+ if (err < 0) {
+ goto out;
+ }
err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
if (err < 0) {
goto out;
@@ -2385,7 +2416,10 @@ static void coroutine_fn v9fs_symlink(void *opaque)
if (err < 0) {
goto out;
}
- stat_to_qid(&stbuf, &qid);
+ err = stat_to_qid(pdu, &stbuf, &qid);
+ if (err < 0) {
+ goto out;
+ }
err = pdu_marshal(pdu, offset, "Q", &qid);
if (err < 0) {
goto out;
@@ -3065,7 +3099,10 @@ static void coroutine_fn v9fs_mknod(void *opaque)
if (err < 0) {
goto out;
}
- stat_to_qid(&stbuf, &qid);
+ err = stat_to_qid(pdu, &stbuf, &qid);
+ if (err < 0) {
+ goto out;
+ }
err = pdu_marshal(pdu, offset, "Q", &qid);
if (err < 0) {
goto out;
@@ -3223,7 +3260,10 @@ static void coroutine_fn v9fs_mkdir(void *opaque)
if (err < 0) {
goto out;
}
- stat_to_qid(&stbuf, &qid);
+ err = stat_to_qid(pdu, &stbuf, &qid);
+ if (err < 0) {
+ goto out;
+ }
err = pdu_marshal(pdu, offset, "Q", &qid);
if (err < 0) {
goto out;
@@ -3634,6 +3674,8 @@ int v9fs_device_realize_common(V9fsState *s, const V9fsTransport *t,
goto out;
}
+ s->dev_id = stat.st_dev;
+
s->ctx.fst = &fse->fst;
fsdev_throttle_init(s->ctx.fst);
diff --git a/hw/9pfs/9p.h b/hw/9pfs/9p.h
index 8883761b2c1d..5e316178d579 100644
--- a/hw/9pfs/9p.h
+++ b/hw/9pfs/9p.h
@@ -256,6 +256,7 @@ struct V9fsState
Error *migration_blocker;
V9fsConf fsconf;
V9fsQID root_qid;
+ dev_t dev_id;
};
/* 9p2000.L open flags */
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PULL v2 5/8] 9p: Added virtfs option 'multidevs=remap|forbid|warn'
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
` (3 preceding siblings ...)
2019-10-10 13:18 ` [PULL v2 4/8] 9p: Treat multiple devices on one export as an error Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-10 13:18 ` [PULL v2 6/8] 9p: stat_to_qid: implement slow path Greg Kurz
` (3 subsequent siblings)
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel
Cc: Peter Maydell, Christian Schoenebeck, Greg Kurz, Antonios Motakis
From: Antonios Motakis <antonios.motakis@huawei.com>
'warn' (default): Only log an error message (once) on host if more than one
device is shared by same export, except of that just ignore this config
error though. This is the default behaviour for not breaking existing
installations implying that they really know what they are doing.
'forbid': Like 'warn', but except of just logging an error this
also denies access of guest to additional devices.
'remap': Allows to share more than one device per export by remapping
inodes from host to guest appropriately. To support multiple devices on the
9p share, and avoid qid path collisions we take the device id as input to
generate a unique QID path. The lowest 48 bits of the path will be set
equal to the file inode, and the top bits will be uniquely assigned based
on the top 16 bits of the inode and the device id.
Signed-off-by: Antonios Motakis <antonios.motakis@huawei.com>
[CS: - Rebased to https://github.com/gkurz/qemu/commits/9p-next
(SHA1 7fc4c49e91).
- Added virtfs option 'multidevs', original patch simply did the inode
remapping without being asked.
- Updated hash calls to new xxhash API.
- Updated docs for new option 'multidevs'.
- Fixed v9fs_do_readdir() not having remapped inodes.
- Log error message when running out of prefixes in
qid_path_prefixmap().
- Fixed definition of QPATH_INO_MASK.
- Wrapped qpp_table initialization to dedicated qpp_table_init()
function.
- Dropped unnecessary parantheses in qpp_lookup_func().
- Dropped unnecessary g_malloc0() result checks. ]
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
[groug: - Moved "multidevs" parsing to the local backend.
- Added hint to invalid multidevs option error.
- Turn "remap" into "x-remap". ]
Signed-off-by: Greg Kurz <groug@kaod.org>
---
fsdev/file-op-9p.h | 5 ++
fsdev/qemu-fsdev-opts.c | 7 +-
fsdev/qemu-fsdev.c | 1 +
hw/9pfs/9p-local.c | 21 +++++
hw/9pfs/9p.c | 188 +++++++++++++++++++++++++++++++++++-----
hw/9pfs/9p.h | 12 +++
qemu-options.hx | 26 +++++-
vl.c | 7 +-
8 files changed, 242 insertions(+), 25 deletions(-)
diff --git a/fsdev/file-op-9p.h b/fsdev/file-op-9p.h
index c757c8099f54..f2f7772c86d5 100644
--- a/fsdev/file-op-9p.h
+++ b/fsdev/file-op-9p.h
@@ -59,6 +59,11 @@ typedef struct ExtendedOps {
#define V9FS_RDONLY 0x00000040
#define V9FS_PROXY_SOCK_FD 0x00000080
#define V9FS_PROXY_SOCK_NAME 0x00000100
+/*
+ * multidevs option (either one of the two applies exclusively)
+ */
+#define V9FS_REMAP_INODES 0x00000200
+#define V9FS_FORBID_MULTIDEVS 0x00000400
#define V9FS_SEC_MASK 0x0000003C
diff --git a/fsdev/qemu-fsdev-opts.c b/fsdev/qemu-fsdev-opts.c
index 7c31ffffafb5..07a18c6e48d3 100644
--- a/fsdev/qemu-fsdev-opts.c
+++ b/fsdev/qemu-fsdev-opts.c
@@ -31,7 +31,9 @@ static QemuOptsList qemu_fsdev_opts = {
}, {
.name = "readonly",
.type = QEMU_OPT_BOOL,
-
+ }, {
+ .name = "multidevs",
+ .type = QEMU_OPT_STRING,
}, {
.name = "socket",
.type = QEMU_OPT_STRING,
@@ -75,6 +77,9 @@ static QemuOptsList qemu_virtfs_opts = {
}, {
.name = "readonly",
.type = QEMU_OPT_BOOL,
+ }, {
+ .name = "multidevs",
+ .type = QEMU_OPT_STRING,
}, {
.name = "socket",
.type = QEMU_OPT_STRING,
diff --git a/fsdev/qemu-fsdev.c b/fsdev/qemu-fsdev.c
index 077a8c4e2bca..a9e069c0c78d 100644
--- a/fsdev/qemu-fsdev.c
+++ b/fsdev/qemu-fsdev.c
@@ -58,6 +58,7 @@ static FsDriverTable FsDrivers[] = {
"writeout",
"fmode",
"dmode",
+ "multidevs",
"throttling.bps-total",
"throttling.bps-read",
"throttling.bps-write",
diff --git a/hw/9pfs/9p-local.c b/hw/9pfs/9p-local.c
index 5c7f4cd9c7b0..4708c0bd896f 100644
--- a/hw/9pfs/9p-local.c
+++ b/hw/9pfs/9p-local.c
@@ -1483,6 +1483,7 @@ static int local_parse_opts(QemuOpts *opts, FsDriverEntry *fse, Error **errp)
{
const char *sec_model = qemu_opt_get(opts, "security_model");
const char *path = qemu_opt_get(opts, "path");
+ const char *multidevs = qemu_opt_get(opts, "multidevs");
Error *local_err = NULL;
if (!sec_model) {
@@ -1506,6 +1507,26 @@ static int local_parse_opts(QemuOpts *opts, FsDriverEntry *fse, Error **errp)
return -1;
}
+ if (multidevs) {
+ if (!strcmp(multidevs, "remap")) {
+ fse->export_flags &= ~V9FS_FORBID_MULTIDEVS;
+ fse->export_flags |= V9FS_REMAP_INODES;
+ } else if (!strcmp(multidevs, "forbid")) {
+ fse->export_flags &= ~V9FS_REMAP_INODES;
+ fse->export_flags |= V9FS_FORBID_MULTIDEVS;
+ } else if (!strcmp(multidevs, "warn")) {
+ fse->export_flags &= ~V9FS_FORBID_MULTIDEVS;
+ fse->export_flags &= ~V9FS_REMAP_INODES;
+ } else {
+ error_setg(&local_err, "invalid multidevs property '%s'",
+ multidevs);
+ error_append_hint(&local_err, "Valid options are: multidevs="
+ "[remap|forbid|warn]\n");
+ error_propagate(errp, local_err);
+ return -1;
+ }
+ }
+
if (!path) {
error_setg(errp, "path property not set");
return -1;
diff --git a/hw/9pfs/9p.c b/hw/9pfs/9p.c
index 5a895ae0bbfe..8eb89c5c7d3f 100644
--- a/hw/9pfs/9p.c
+++ b/hw/9pfs/9p.c
@@ -26,6 +26,7 @@
#include "trace.h"
#include "migration/blocker.h"
#include "sysemu/qtest.h"
+#include "qemu/xxhash.h"
int open_fd_hw;
int total_open_fd;
@@ -572,23 +573,117 @@ static void coroutine_fn virtfs_reset(V9fsPDU *pdu)
P9_STAT_MODE_NAMED_PIPE | \
P9_STAT_MODE_SOCKET)
-/* This is the algorithm from ufs in spfs */
+/* creative abuse of tb_hash_func7, which is based on xxhash */
+static uint32_t qpp_hash(QppEntry e)
+{
+ return qemu_xxhash7(e.ino_prefix, e.dev, 0, 0, 0);
+}
+
+static bool qpp_lookup_func(const void *obj, const void *userp)
+{
+ const QppEntry *e1 = obj, *e2 = userp;
+ return e1->dev == e2->dev && e1->ino_prefix == e2->ino_prefix;
+}
+
+static void qpp_table_remove(void *p, uint32_t h, void *up)
+{
+ g_free(p);
+}
+
+static void qpp_table_destroy(struct qht *ht)
+{
+ if (!ht || !ht->map) {
+ return;
+ }
+ qht_iter(ht, qpp_table_remove, NULL);
+ qht_destroy(ht);
+}
+
+static void qpp_table_init(struct qht *ht)
+{
+ qht_init(ht, qpp_lookup_func, 1, QHT_MODE_AUTO_RESIZE);
+}
+
+/*
+ * stat_to_qid needs to map inode number (64 bits) and device id (32 bits)
+ * to a unique QID path (64 bits). To avoid having to map and keep track
+ * of up to 2^64 objects, we map only the 16 highest bits of the inode plus
+ * the device id to the 16 highest bits of the QID path. The 48 lowest bits
+ * of the QID path equal to the lowest bits of the inode number.
+ *
+ * This takes advantage of the fact that inode number are usually not
+ * random but allocated sequentially, so we have fewer items to keep
+ * track of.
+ */
+static int qid_path_prefixmap(V9fsPDU *pdu, const struct stat *stbuf,
+ uint64_t *path)
+{
+ QppEntry lookup = {
+ .dev = stbuf->st_dev,
+ .ino_prefix = (uint16_t) (stbuf->st_ino >> 48)
+ }, *val;
+ uint32_t hash = qpp_hash(lookup);
+
+ val = qht_lookup(&pdu->s->qpp_table, &lookup, hash);
+
+ if (!val) {
+ if (pdu->s->qp_prefix_next == 0) {
+ /* we ran out of prefixes */
+ error_report_once(
+ "9p: No more prefixes available for remapping inodes from "
+ "host to guest."
+ );
+ return -ENFILE;
+ }
+
+ val = g_malloc0(sizeof(QppEntry));
+ *val = lookup;
+
+ /* new unique inode prefix and device combo */
+ val->qp_prefix = pdu->s->qp_prefix_next++;
+ qht_insert(&pdu->s->qpp_table, val, hash, NULL);
+ }
+
+ *path = ((uint64_t)val->qp_prefix << 48) | (stbuf->st_ino & QPATH_INO_MASK);
+ return 0;
+}
+
static int stat_to_qid(V9fsPDU *pdu, const struct stat *stbuf, V9fsQID *qidp)
{
+ int err;
size_t size;
- if (pdu->s->dev_id != stbuf->st_dev) {
- warn_report_once(
- "9p: Multiple devices detected in same VirtFS export, "
- "which might lead to file ID collisions and severe "
- "misbehaviours on guest! You should use a separate "
- "export for each device shared from host."
- );
+ if (pdu->s->ctx.export_flags & V9FS_REMAP_INODES) {
+ /* map inode+device to qid path (fast path) */
+ err = qid_path_prefixmap(pdu, stbuf, &qidp->path);
+ if (err) {
+ return err;
+ }
+ } else {
+ if (pdu->s->dev_id != stbuf->st_dev) {
+ if (pdu->s->ctx.export_flags & V9FS_FORBID_MULTIDEVS) {
+ error_report_once(
+ "9p: Multiple devices detected in same VirtFS export. "
+ "Access of guest to additional devices is (partly) "
+ "denied due to virtfs option 'multidevs=forbid' being "
+ "effective."
+ );
+ return -ENODEV;
+ } else {
+ warn_report_once(
+ "9p: Multiple devices detected in same VirtFS export, "
+ "which might lead to file ID collisions and severe "
+ "misbehaviours on guest! You should either use a "
+ "separate export for each device shared from host or "
+ "use virtfs option 'multidevs=remap'!"
+ );
+ }
+ }
+ memset(&qidp->path, 0, sizeof(qidp->path));
+ size = MIN(sizeof(stbuf->st_ino), sizeof(qidp->path));
+ memcpy(&qidp->path, &stbuf->st_ino, size);
}
- memset(&qidp->path, 0, sizeof(qidp->path));
- size = MIN(sizeof(stbuf->st_ino), sizeof(qidp->path));
- memcpy(&qidp->path, &stbuf->st_ino, size);
qidp->version = stbuf->st_mtime ^ (stbuf->st_size << 8);
qidp->type = 0;
if (S_ISDIR(stbuf->st_mode)) {
@@ -618,6 +713,30 @@ static int coroutine_fn fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp,
return 0;
}
+static int coroutine_fn dirent_to_qid(V9fsPDU *pdu, V9fsFidState *fidp,
+ struct dirent *dent, V9fsQID *qidp)
+{
+ struct stat stbuf;
+ V9fsPath path;
+ int err;
+
+ v9fs_path_init(&path);
+
+ err = v9fs_co_name_to_path(pdu, &fidp->path, dent->d_name, &path);
+ if (err < 0) {
+ goto out;
+ }
+ err = v9fs_co_lstat(pdu, &path, &stbuf);
+ if (err < 0) {
+ goto out;
+ }
+ err = stat_to_qid(pdu, &stbuf, qidp);
+
+out:
+ v9fs_path_free(&path);
+ return err;
+}
+
V9fsPDU *pdu_alloc(V9fsState *s)
{
V9fsPDU *pdu = NULL;
@@ -1966,16 +2085,39 @@ static int coroutine_fn v9fs_do_readdir(V9fsPDU *pdu, V9fsFidState *fidp,
v9fs_string_free(&name);
return count;
}
- /*
- * Fill up just the path field of qid because the client uses
- * only that. To fill the entire qid structure we will have
- * to stat each dirent found, which is expensive
- */
- size = MIN(sizeof(dent->d_ino), sizeof(qid.path));
- memcpy(&qid.path, &dent->d_ino, size);
- /* Fill the other fields with dummy values */
- qid.type = 0;
- qid.version = 0;
+
+ if (pdu->s->ctx.export_flags & V9FS_REMAP_INODES) {
+ /*
+ * dirent_to_qid() implies expensive stat call for each entry,
+ * we must do that here though since inode remapping requires
+ * the device id, which in turn might be different for
+ * different entries; we cannot make any assumption to avoid
+ * that here.
+ */
+ err = dirent_to_qid(pdu, fidp, dent, &qid);
+ if (err < 0) {
+ v9fs_readdir_unlock(&fidp->fs.dir);
+ v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
+ v9fs_string_free(&name);
+ return err;
+ }
+ } else {
+ /*
+ * Fill up just the path field of qid because the client uses
+ * only that. To fill the entire qid structure we will have
+ * to stat each dirent found, which is expensive. For the
+ * latter reason we don't call dirent_to_qid() here. Only drawback
+ * is that no multi-device export detection of stat_to_qid()
+ * would be done and provided as error to the user here. But
+ * user would get that error anyway when accessing those
+ * files/dirs through other ways.
+ */
+ size = MIN(sizeof(dent->d_ino), sizeof(qid.path));
+ memcpy(&qid.path, &dent->d_ino, size);
+ /* Fill the other fields with dummy values */
+ qid.type = 0;
+ qid.version = 0;
+ }
/* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
len = pdu_marshal(pdu, 11 + count, "Qqbs",
@@ -3676,6 +3818,9 @@ int v9fs_device_realize_common(V9fsState *s, const V9fsTransport *t,
s->dev_id = stat.st_dev;
+ qpp_table_init(&s->qpp_table);
+ s->qp_prefix_next = 1; /* reserve 0 to detect overflow */
+
s->ctx.fst = &fse->fst;
fsdev_throttle_init(s->ctx.fst);
@@ -3697,6 +3842,7 @@ void v9fs_device_unrealize_common(V9fsState *s, Error **errp)
fsdev_throttle_cleanup(s->ctx.fst);
}
g_free(s->tag);
+ qpp_table_destroy(&s->qpp_table);
g_free(s->ctx.fs_root);
}
diff --git a/hw/9pfs/9p.h b/hw/9pfs/9p.h
index 5e316178d579..7262fe80aa02 100644
--- a/hw/9pfs/9p.h
+++ b/hw/9pfs/9p.h
@@ -8,6 +8,7 @@
#include "fsdev/9p-iov-marshal.h"
#include "qemu/thread.h"
#include "qemu/coroutine.h"
+#include "qemu/qht.h"
enum {
P9_TLERROR = 6,
@@ -235,6 +236,15 @@ struct V9fsFidState
V9fsFidState *rclm_lst;
};
+#define QPATH_INO_MASK ((1ULL << 48) - 1)
+
+/* QID path prefix entry, see stat_to_qid */
+typedef struct {
+ dev_t dev;
+ uint16_t ino_prefix;
+ uint16_t qp_prefix;
+} QppEntry;
+
struct V9fsState
{
QLIST_HEAD(, V9fsPDU) free_list;
@@ -257,6 +267,8 @@ struct V9fsState
V9fsConf fsconf;
V9fsQID root_qid;
dev_t dev_id;
+ struct qht qpp_table;
+ uint16_t qp_prefix_next;
};
/* 9p2000.L open flags */
diff --git a/qemu-options.hx b/qemu-options.hx
index 2a04ca6ac509..793d70ff9388 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -1339,7 +1339,7 @@ ETEXI
DEF("virtfs", HAS_ARG, QEMU_OPTION_virtfs,
"-virtfs local,path=path,mount_tag=tag,security_model=mapped-xattr|mapped-file|passthrough|none\n"
- " [,id=id][,writeout=immediate][,readonly][,fmode=fmode][,dmode=dmode]\n"
+ " [,id=id][,writeout=immediate][,readonly][,fmode=fmode][,dmode=dmode][,multidevs=remap|forbid|warn]\n"
"-virtfs proxy,mount_tag=tag,socket=socket[,id=id][,writeout=immediate][,readonly]\n"
"-virtfs proxy,mount_tag=tag,sock_fd=sock_fd[,id=id][,writeout=immediate][,readonly]\n"
"-virtfs synth,mount_tag=tag[,id=id][,readonly]\n",
@@ -1347,7 +1347,7 @@ DEF("virtfs", HAS_ARG, QEMU_OPTION_virtfs,
STEXI
-@item -virtfs local,path=@var{path},mount_tag=@var{mount_tag} ,security_model=@var{security_model}[,writeout=@var{writeout}][,readonly] [,fmode=@var{fmode}][,dmode=@var{dmode}]
+@item -virtfs local,path=@var{path},mount_tag=@var{mount_tag} ,security_model=@var{security_model}[,writeout=@var{writeout}][,readonly] [,fmode=@var{fmode}][,dmode=@var{dmode}][,multidevs=@var{multidevs}]
@itemx -virtfs proxy,socket=@var{socket},mount_tag=@var{mount_tag} [,writeout=@var{writeout}][,readonly]
@itemx -virtfs proxy,sock_fd=@var{sock_fd},mount_tag=@var{mount_tag} [,writeout=@var{writeout}][,readonly]
@itemx -virtfs synth,mount_tag=@var{mount_tag}
@@ -1403,6 +1403,28 @@ Specifies the default mode for newly created directories on the host. Works
only with security models "mapped-xattr" and "mapped-file".
@item mount_tag=@var{mount_tag}
Specifies the tag name to be used by the guest to mount this export point.
+@item multidevs=@var{multidevs}
+Specifies how to deal with multiple devices being shared with a 9p export.
+Supported behaviours are either "remap", "forbid" or "warn". The latter is
+the default behaviour on which virtfs 9p expects only one device to be
+shared with the same export, and if more than one device is shared and
+accessed via the same 9p export then only a warning message is logged
+(once) by qemu on host side. In order to avoid file ID collisions on guest
+you should either create a separate virtfs export for each device to be
+shared with guests (recommended way) or you might use "remap" instead which
+allows you to share multiple devices with only one export instead, which is
+achieved by remapping the original inode numbers from host to guest in a
+way that would prevent such collisions. Remapping inodes in such use cases
+is required because the original device IDs from host are never passed and
+exposed on guest. Instead all files of an export shared with virtfs always
+share the same device id on guest. So two files with identical inode
+numbers but from actually different devices on host would otherwise cause a
+file ID collision and hence potential misbehaviours on guest. "forbid" on
+the other hand assumes like "warn" that only one device is shared by the
+same export, however it will not only log a warning message but also
+deny access to additional devices on guest. Note though that "forbid" does
+currently not block all possible file access operations (e.g. readdir()
+would still return entries from other devices).
@end table
ETEXI
diff --git a/vl.c b/vl.c
index 002bf4919e1e..0a295e5d77d6 100644
--- a/vl.c
+++ b/vl.c
@@ -3335,7 +3335,8 @@ int main(int argc, char **argv, char **envp)
case QEMU_OPTION_virtfs: {
QemuOpts *fsdev;
QemuOpts *device;
- const char *writeout, *sock_fd, *socket, *path, *security_model;
+ const char *writeout, *sock_fd, *socket, *path, *security_model,
+ *multidevs;
olist = qemu_find_opts("virtfs");
if (!olist) {
@@ -3395,6 +3396,10 @@ int main(int argc, char **argv, char **envp)
qemu_opt_set_bool(fsdev, "readonly",
qemu_opt_get_bool(opts, "readonly", 0),
&error_abort);
+ multidevs = qemu_opt_get(opts, "multidevs");
+ if (multidevs) {
+ qemu_opt_set(fsdev, "multidevs", multidevs, &error_abort);
+ }
device = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
&error_abort);
qemu_opt_set(device, "driver", "virtio-9p-pci", &error_abort);
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PULL v2 6/8] 9p: stat_to_qid: implement slow path
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
` (4 preceding siblings ...)
2019-10-10 13:18 ` [PULL v2 5/8] 9p: Added virtfs option 'multidevs=remap|forbid|warn' Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-10 13:18 ` [PULL v2 7/8] 9p: Use variable length suffixes for inode remapping Greg Kurz
` (2 subsequent siblings)
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel
Cc: Peter Maydell, Christian Schoenebeck, Greg Kurz, Antonios Motakis
From: Antonios Motakis <antonios.motakis@huawei.com>
stat_to_qid attempts via qid_path_prefixmap to map unique files (which are
identified by 64 bit inode nr and 32 bit device id) to a 64 QID path value.
However this implementation makes some assumptions about inode number
generation on the host.
If qid_path_prefixmap fails, we still have 48 bits available in the QID
path to fall back to a less memory efficient full mapping.
Signed-off-by: Antonios Motakis <antonios.motakis@huawei.com>
[CS: - Rebased to https://github.com/gkurz/qemu/commits/9p-next
(SHA1 7fc4c49e91).
- Updated hash calls to new xxhash API.
- Removed unnecessary parantheses in qpf_lookup_func().
- Removed unnecessary g_malloc0() result checks.
- Log error message when running out of prefixes in
qid_path_fullmap().
- Log warning message about potential degraded performance in
qid_path_prefixmap().
- Wrapped qpf_table initialization to dedicated qpf_table_init()
function.
- Fixed typo in comment. ]
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
---
hw/9pfs/9p.c | 74 +++++++++++++++++++++++++++++++++++++++++++++++-----
hw/9pfs/9p.h | 9 +++++++
2 files changed, 76 insertions(+), 7 deletions(-)
diff --git a/hw/9pfs/9p.c b/hw/9pfs/9p.c
index 8eb89c5c7d3f..d9be2d45d3e4 100644
--- a/hw/9pfs/9p.c
+++ b/hw/9pfs/9p.c
@@ -579,23 +579,34 @@ static uint32_t qpp_hash(QppEntry e)
return qemu_xxhash7(e.ino_prefix, e.dev, 0, 0, 0);
}
+static uint32_t qpf_hash(QpfEntry e)
+{
+ return qemu_xxhash7(e.ino, e.dev, 0, 0, 0);
+}
+
static bool qpp_lookup_func(const void *obj, const void *userp)
{
const QppEntry *e1 = obj, *e2 = userp;
return e1->dev == e2->dev && e1->ino_prefix == e2->ino_prefix;
}
-static void qpp_table_remove(void *p, uint32_t h, void *up)
+static bool qpf_lookup_func(const void *obj, const void *userp)
+{
+ const QpfEntry *e1 = obj, *e2 = userp;
+ return e1->dev == e2->dev && e1->ino == e2->ino;
+}
+
+static void qp_table_remove(void *p, uint32_t h, void *up)
{
g_free(p);
}
-static void qpp_table_destroy(struct qht *ht)
+static void qp_table_destroy(struct qht *ht)
{
if (!ht || !ht->map) {
return;
}
- qht_iter(ht, qpp_table_remove, NULL);
+ qht_iter(ht, qp_table_remove, NULL);
qht_destroy(ht);
}
@@ -604,6 +615,50 @@ static void qpp_table_init(struct qht *ht)
qht_init(ht, qpp_lookup_func, 1, QHT_MODE_AUTO_RESIZE);
}
+static void qpf_table_init(struct qht *ht)
+{
+ qht_init(ht, qpf_lookup_func, 1 << 16, QHT_MODE_AUTO_RESIZE);
+}
+
+static int qid_path_fullmap(V9fsPDU *pdu, const struct stat *stbuf,
+ uint64_t *path)
+{
+ QpfEntry lookup = {
+ .dev = stbuf->st_dev,
+ .ino = stbuf->st_ino
+ }, *val;
+ uint32_t hash = qpf_hash(lookup);
+
+ /* most users won't need the fullmap, so init the table lazily */
+ if (!pdu->s->qpf_table.map) {
+ qpf_table_init(&pdu->s->qpf_table);
+ }
+
+ val = qht_lookup(&pdu->s->qpf_table, &lookup, hash);
+
+ if (!val) {
+ if (pdu->s->qp_fullpath_next == 0) {
+ /* no more files can be mapped :'( */
+ error_report_once(
+ "9p: No more prefixes available for remapping inodes from "
+ "host to guest."
+ );
+ return -ENFILE;
+ }
+
+ val = g_malloc0(sizeof(QppEntry));
+ *val = lookup;
+
+ /* new unique inode and device combo */
+ val->path = pdu->s->qp_fullpath_next++;
+ pdu->s->qp_fullpath_next &= QPATH_INO_MASK;
+ qht_insert(&pdu->s->qpf_table, val, hash, NULL);
+ }
+
+ *path = val->path;
+ return 0;
+}
+
/*
* stat_to_qid needs to map inode number (64 bits) and device id (32 bits)
* to a unique QID path (64 bits). To avoid having to map and keep track
@@ -629,9 +684,8 @@ static int qid_path_prefixmap(V9fsPDU *pdu, const struct stat *stbuf,
if (!val) {
if (pdu->s->qp_prefix_next == 0) {
/* we ran out of prefixes */
- error_report_once(
- "9p: No more prefixes available for remapping inodes from "
- "host to guest."
+ warn_report_once(
+ "9p: Potential degraded performance of inode remapping"
);
return -ENFILE;
}
@@ -656,6 +710,10 @@ static int stat_to_qid(V9fsPDU *pdu, const struct stat *stbuf, V9fsQID *qidp)
if (pdu->s->ctx.export_flags & V9FS_REMAP_INODES) {
/* map inode+device to qid path (fast path) */
err = qid_path_prefixmap(pdu, stbuf, &qidp->path);
+ if (err == -ENFILE) {
+ /* fast path didn't work, fall back to full map */
+ err = qid_path_fullmap(pdu, stbuf, &qidp->path);
+ }
if (err) {
return err;
}
@@ -3820,6 +3878,7 @@ int v9fs_device_realize_common(V9fsState *s, const V9fsTransport *t,
qpp_table_init(&s->qpp_table);
s->qp_prefix_next = 1; /* reserve 0 to detect overflow */
+ s->qp_fullpath_next = 1;
s->ctx.fst = &fse->fst;
fsdev_throttle_init(s->ctx.fst);
@@ -3842,7 +3901,8 @@ void v9fs_device_unrealize_common(V9fsState *s, Error **errp)
fsdev_throttle_cleanup(s->ctx.fst);
}
g_free(s->tag);
- qpp_table_destroy(&s->qpp_table);
+ qp_table_destroy(&s->qpp_table);
+ qp_table_destroy(&s->qpf_table);
g_free(s->ctx.fs_root);
}
diff --git a/hw/9pfs/9p.h b/hw/9pfs/9p.h
index 7262fe80aa02..35a362c0d797 100644
--- a/hw/9pfs/9p.h
+++ b/hw/9pfs/9p.h
@@ -245,6 +245,13 @@ typedef struct {
uint16_t qp_prefix;
} QppEntry;
+/* QID path full entry, as above */
+typedef struct {
+ dev_t dev;
+ ino_t ino;
+ uint64_t path;
+} QpfEntry;
+
struct V9fsState
{
QLIST_HEAD(, V9fsPDU) free_list;
@@ -268,7 +275,9 @@ struct V9fsState
V9fsQID root_qid;
dev_t dev_id;
struct qht qpp_table;
+ struct qht qpf_table;
uint16_t qp_prefix_next;
+ uint64_t qp_fullpath_next;
};
/* 9p2000.L open flags */
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PULL v2 7/8] 9p: Use variable length suffixes for inode remapping
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
` (5 preceding siblings ...)
2019-10-10 13:18 ` [PULL v2 6/8] 9p: stat_to_qid: implement slow path Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-10 13:18 ` [PULL v2 8/8] MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes" Greg Kurz
2019-10-14 14:08 ` [PULL v2 0/8] 9p patches 2019-10-10 Peter Maydell
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel; +Cc: Peter Maydell, Christian Schoenebeck, Greg Kurz
From: Christian Schoenebeck <qemu_oss@crudebyte.com>
Use variable length suffixes for inode remapping instead of the fixed
16 bit size prefixes before. With this change the inode numbers on guest
will typically be much smaller (e.g. around >2^1 .. >2^7 instead of >2^48
with the previous fixed size inode remapping.
Additionally this solution is more efficient, since inode numbers in
practice can take almost their entire 64 bit range on guest as well, so
there is less likely a need for generating and tracking additional suffixes,
which might also be beneficial for nested virtualization where each level of
virtualization would shift up the inode bits and increase the chance of
expensive remapping actions.
The "Exponential Golomb" algorithm is used as basis for generating the
variable length suffixes. The algorithm has a parameter k which controls the
distribution of bits on increasing indeces (minimum bits at low index vs.
maximum bits at high index). With k=0 the generated suffixes look like:
Index Dec/Bin -> Generated Suffix Bin
1 [1] -> [1] (1 bits)
2 [10] -> [010] (3 bits)
3 [11] -> [110] (3 bits)
4 [100] -> [00100] (5 bits)
5 [101] -> [10100] (5 bits)
6 [110] -> [01100] (5 bits)
7 [111] -> [11100] (5 bits)
8 [1000] -> [0001000] (7 bits)
9 [1001] -> [1001000] (7 bits)
10 [1010] -> [0101000] (7 bits)
11 [1011] -> [1101000] (7 bits)
12 [1100] -> [0011000] (7 bits)
...
65533 [1111111111111101] -> [1011111111111111000000000000000] (31 bits)
65534 [1111111111111110] -> [0111111111111111000000000000000] (31 bits)
65535 [1111111111111111] -> [1111111111111111000000000000000] (31 bits)
Hence minBits=1 maxBits=31
And with k=5 they would look like:
Index Dec/Bin -> Generated Suffix Bin
1 [1] -> [000001] (6 bits)
2 [10] -> [100001] (6 bits)
3 [11] -> [010001] (6 bits)
4 [100] -> [110001] (6 bits)
5 [101] -> [001001] (6 bits)
6 [110] -> [101001] (6 bits)
7 [111] -> [011001] (6 bits)
8 [1000] -> [111001] (6 bits)
9 [1001] -> [000101] (6 bits)
10 [1010] -> [100101] (6 bits)
11 [1011] -> [010101] (6 bits)
12 [1100] -> [110101] (6 bits)
...
65533 [1111111111111101] -> [0011100000000000100000000000] (28 bits)
65534 [1111111111111110] -> [1011100000000000100000000000] (28 bits)
65535 [1111111111111111] -> [0111100000000000100000000000] (28 bits)
Hence minBits=6 maxBits=28
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
---
hw/9pfs/9p.c | 268 +++++++++++++++++++++++++++++++++++++++++++++------
hw/9pfs/9p.h | 44 ++++++++-
2 files changed, 279 insertions(+), 33 deletions(-)
diff --git a/hw/9pfs/9p.c b/hw/9pfs/9p.c
index d9be2d45d3e4..37abcdb71ec0 100644
--- a/hw/9pfs/9p.c
+++ b/hw/9pfs/9p.c
@@ -27,6 +27,7 @@
#include "migration/blocker.h"
#include "sysemu/qtest.h"
#include "qemu/xxhash.h"
+#include <math.h>
int open_fd_hw;
int total_open_fd;
@@ -573,6 +574,116 @@ static void coroutine_fn virtfs_reset(V9fsPDU *pdu)
P9_STAT_MODE_NAMED_PIPE | \
P9_STAT_MODE_SOCKET)
+/* Mirrors all bits of a byte. So e.g. binary 10100000 would become 00000101. */
+static inline uint8_t mirror8bit(uint8_t byte)
+{
+ return (byte * 0x0202020202ULL & 0x010884422010ULL) % 1023;
+}
+
+/* Same as mirror8bit() just for a 64 bit data type instead for a byte. */
+static inline uint64_t mirror64bit(uint64_t value)
+{
+ return ((uint64_t)mirror8bit(value & 0xff) << 56) |
+ ((uint64_t)mirror8bit((value >> 8) & 0xff) << 48) |
+ ((uint64_t)mirror8bit((value >> 16) & 0xff) << 40) |
+ ((uint64_t)mirror8bit((value >> 24) & 0xff) << 32) |
+ ((uint64_t)mirror8bit((value >> 32) & 0xff) << 24) |
+ ((uint64_t)mirror8bit((value >> 40) & 0xff) << 16) |
+ ((uint64_t)mirror8bit((value >> 48) & 0xff) << 8) |
+ ((uint64_t)mirror8bit((value >> 56) & 0xff));
+}
+
+/**
+ * @brief Parameter k for the Exponential Golomb algorihm to be used.
+ *
+ * The smaller this value, the smaller the minimum bit count for the Exp.
+ * Golomb generated affixes will be (at lowest index) however for the
+ * price of having higher maximum bit count of generated affixes (at highest
+ * index). Likewise increasing this parameter yields in smaller maximum bit
+ * count for the price of having higher minimum bit count.
+ *
+ * In practice that means: a good value for k depends on the expected amount
+ * of devices to be exposed by one export. For a small amount of devices k
+ * should be small, for a large amount of devices k might be increased
+ * instead. The default of k=0 should be fine for most users though.
+ *
+ * @b IMPORTANT: In case this ever becomes a runtime parameter; the value of
+ * k should not change as long as guest is still running! Because that would
+ * cause completely different inode numbers to be generated on guest.
+ */
+#define EXP_GOLOMB_K 0
+
+/**
+ * @brief Exponential Golomb algorithm for arbitrary k (including k=0).
+ *
+ * The Exponential Golomb algorithm generates @b prefixes (@b not suffixes!)
+ * with growing length and with the mathematical property of being
+ * "prefix-free". The latter means the generated prefixes can be prepended
+ * in front of arbitrary numbers and the resulting concatenated numbers are
+ * guaranteed to be always unique.
+ *
+ * This is a minor adjustment to the original Exp. Golomb algorithm in the
+ * sense that lowest allowed index (@param n) starts with 1, not with zero.
+ *
+ * @param n - natural number (or index) of the prefix to be generated
+ * (1, 2, 3, ...)
+ * @param k - parameter k of Exp. Golomb algorithm to be used
+ * (see comment on EXP_GOLOMB_K macro for details about k)
+ */
+static VariLenAffix expGolombEncode(uint64_t n, int k)
+{
+ const uint64_t value = n + (1 << k) - 1;
+ const int bits = (int) log2(value) + 1;
+ return (VariLenAffix) {
+ .type = AffixType_Prefix,
+ .value = value,
+ .bits = bits + MAX((bits - 1 - k), 0)
+ };
+}
+
+/**
+ * @brief Converts a suffix into a prefix, or a prefix into a suffix.
+ *
+ * Simply mirror all bits of the affix value, for the purpose to preserve
+ * respectively the mathematical "prefix-free" or "suffix-free" property
+ * after the conversion.
+ *
+ * If a passed prefix is suitable to create unique numbers, then the
+ * returned suffix is suitable to create unique numbers as well (and vice
+ * versa).
+ */
+static VariLenAffix invertAffix(const VariLenAffix *affix)
+{
+ return (VariLenAffix) {
+ .type =
+ (affix->type == AffixType_Suffix) ?
+ AffixType_Prefix : AffixType_Suffix,
+ .value =
+ mirror64bit(affix->value) >>
+ ((sizeof(affix->value) * 8) - affix->bits),
+ .bits = affix->bits
+ };
+}
+
+/**
+ * @brief Generates suffix numbers with "suffix-free" property.
+ *
+ * This is just a wrapper function on top of the Exp. Golomb algorithm.
+ *
+ * Since the Exp. Golomb algorithm generates prefixes, but we need suffixes,
+ * this function converts the Exp. Golomb prefixes into appropriate suffixes
+ * which are still suitable for generating unique numbers.
+ *
+ * @param n - natural number (or index) of the suffix to be generated
+ * (1, 2, 3, ...)
+ */
+static VariLenAffix affixForIndex(uint64_t index)
+{
+ VariLenAffix prefix;
+ prefix = expGolombEncode(index, EXP_GOLOMB_K);
+ return invertAffix(&prefix); /* convert prefix to suffix */
+}
+
/* creative abuse of tb_hash_func7, which is based on xxhash */
static uint32_t qpp_hash(QppEntry e)
{
@@ -584,13 +695,19 @@ static uint32_t qpf_hash(QpfEntry e)
return qemu_xxhash7(e.ino, e.dev, 0, 0, 0);
}
-static bool qpp_lookup_func(const void *obj, const void *userp)
+static bool qpd_cmp_func(const void *obj, const void *userp)
+{
+ const QpdEntry *e1 = obj, *e2 = userp;
+ return e1->dev == e2->dev;
+}
+
+static bool qpp_cmp_func(const void *obj, const void *userp)
{
const QppEntry *e1 = obj, *e2 = userp;
return e1->dev == e2->dev && e1->ino_prefix == e2->ino_prefix;
}
-static bool qpf_lookup_func(const void *obj, const void *userp)
+static bool qpf_cmp_func(const void *obj, const void *userp)
{
const QpfEntry *e1 = obj, *e2 = userp;
return e1->dev == e2->dev && e1->ino == e2->ino;
@@ -610,16 +727,70 @@ static void qp_table_destroy(struct qht *ht)
qht_destroy(ht);
}
+static void qpd_table_init(struct qht *ht)
+{
+ qht_init(ht, qpd_cmp_func, 1, QHT_MODE_AUTO_RESIZE);
+}
+
static void qpp_table_init(struct qht *ht)
{
- qht_init(ht, qpp_lookup_func, 1, QHT_MODE_AUTO_RESIZE);
+ qht_init(ht, qpp_cmp_func, 1, QHT_MODE_AUTO_RESIZE);
}
static void qpf_table_init(struct qht *ht)
{
- qht_init(ht, qpf_lookup_func, 1 << 16, QHT_MODE_AUTO_RESIZE);
+ qht_init(ht, qpf_cmp_func, 1 << 16, QHT_MODE_AUTO_RESIZE);
+}
+
+/*
+ * Returns how many (high end) bits of inode numbers of the passed fs
+ * device shall be used (in combination with the device number) to
+ * generate hash values for qpp_table entries.
+ *
+ * This function is required if variable length suffixes are used for inode
+ * number mapping on guest level. Since a device may end up having multiple
+ * entries in qpp_table, each entry most probably with a different suffix
+ * length, we thus need this function in conjunction with qpd_table to
+ * "agree" about a fix amount of bits (per device) to be always used for
+ * generating hash values for the purpose of accessing qpp_table in order
+ * get consistent behaviour when accessing qpp_table.
+ */
+static int qid_inode_prefix_hash_bits(V9fsPDU *pdu, dev_t dev)
+{
+ QpdEntry lookup = {
+ .dev = dev
+ }, *val;
+ uint32_t hash = dev;
+ VariLenAffix affix;
+
+ val = qht_lookup(&pdu->s->qpd_table, &lookup, hash);
+ if (!val) {
+ val = g_malloc0(sizeof(QpdEntry));
+ *val = lookup;
+ affix = affixForIndex(pdu->s->qp_affix_next);
+ val->prefix_bits = affix.bits;
+ qht_insert(&pdu->s->qpd_table, val, hash, NULL);
+ pdu->s->qp_ndevices++;
+ }
+ return val->prefix_bits;
}
+/**
+ * @brief Slow / full mapping host inode nr -> guest inode nr.
+ *
+ * This function performs a slower and much more costly remapping of an
+ * original file inode number on host to an appropriate different inode
+ * number on guest. For every (dev, inode) combination on host a new
+ * sequential number is generated, cached and exposed as inode number on
+ * guest.
+ *
+ * This is just a "last resort" fallback solution if the much faster/cheaper
+ * qid_path_suffixmap() failed. In practice this slow / full mapping is not
+ * expected ever to be used at all though.
+ *
+ * @see qid_path_suffixmap() for details
+ *
+ */
static int qid_path_fullmap(V9fsPDU *pdu, const struct stat *stbuf,
uint64_t *path)
{
@@ -628,11 +799,7 @@ static int qid_path_fullmap(V9fsPDU *pdu, const struct stat *stbuf,
.ino = stbuf->st_ino
}, *val;
uint32_t hash = qpf_hash(lookup);
-
- /* most users won't need the fullmap, so init the table lazily */
- if (!pdu->s->qpf_table.map) {
- qpf_table_init(&pdu->s->qpf_table);
- }
+ VariLenAffix affix;
val = qht_lookup(&pdu->s->qpf_table, &lookup, hash);
@@ -650,8 +817,11 @@ static int qid_path_fullmap(V9fsPDU *pdu, const struct stat *stbuf,
*val = lookup;
/* new unique inode and device combo */
- val->path = pdu->s->qp_fullpath_next++;
- pdu->s->qp_fullpath_next &= QPATH_INO_MASK;
+ affix = affixForIndex(
+ 1ULL << (sizeof(pdu->s->qp_affix_next) * 8)
+ );
+ val->path = (pdu->s->qp_fullpath_next++ << affix.bits) | affix.value;
+ pdu->s->qp_fullpath_next &= ((1ULL << (64 - affix.bits)) - 1);
qht_insert(&pdu->s->qpf_table, val, hash, NULL);
}
@@ -659,31 +829,60 @@ static int qid_path_fullmap(V9fsPDU *pdu, const struct stat *stbuf,
return 0;
}
-/*
- * stat_to_qid needs to map inode number (64 bits) and device id (32 bits)
- * to a unique QID path (64 bits). To avoid having to map and keep track
- * of up to 2^64 objects, we map only the 16 highest bits of the inode plus
- * the device id to the 16 highest bits of the QID path. The 48 lowest bits
- * of the QID path equal to the lowest bits of the inode number.
+/**
+ * @brief Quick mapping host inode nr -> guest inode nr.
*
- * This takes advantage of the fact that inode number are usually not
- * random but allocated sequentially, so we have fewer items to keep
- * track of.
+ * This function performs quick remapping of an original file inode number
+ * on host to an appropriate different inode number on guest. This remapping
+ * of inodes is required to avoid inode nr collisions on guest which would
+ * happen if the 9p export contains more than 1 exported file system (or
+ * more than 1 file system data set), because unlike on host level where the
+ * files would have different device nrs, all files exported by 9p would
+ * share the same device nr on guest (the device nr of the virtual 9p device
+ * that is).
+ *
+ * Inode remapping is performed by chopping off high end bits of the original
+ * inode number from host, shifting the result upwards and then assigning a
+ * generated suffix number for the low end bits, where the same suffix number
+ * will be shared by all inodes with the same device id AND the same high end
+ * bits that have been chopped off. That approach utilizes the fact that inode
+ * numbers very likely share the same high end bits (i.e. due to their common
+ * sequential generation by file systems) and hence we only have to generate
+ * and track a very limited amount of suffixes in practice due to that.
+ *
+ * We generate variable size suffixes for that purpose. The 1st generated
+ * suffix will only have 1 bit and hence we only need to chop off 1 bit from
+ * the original inode number. The subsequent suffixes being generated will
+ * grow in (bit) size subsequently, i.e. the 2nd and 3rd suffix being
+ * generated will have 3 bits and hence we have to chop off 3 bits from their
+ * original inodes, and so on. That approach of using variable length suffixes
+ * (i.e. over fixed size ones) utilizes the fact that in practice only a very
+ * limited amount of devices are shared by the same export (e.g. typically
+ * less than 2 dozen devices per 9p export), so in practice we need to chop
+ * off less bits than with fixed size prefixes and yet are flexible to add
+ * new devices at runtime below host's export directory at any time without
+ * having to reboot guest nor requiring to reconfigure guest for that. And due
+ * to the very limited amount of original high end bits that we chop off that
+ * way, the total amount of suffixes we need to generate is less than by using
+ * fixed size prefixes and hence it also improves performance of the inode
+ * remapping algorithm, and finally has the nice side effect that the inode
+ * numbers on guest will be much smaller & human friendly. ;-)
*/
-static int qid_path_prefixmap(V9fsPDU *pdu, const struct stat *stbuf,
+static int qid_path_suffixmap(V9fsPDU *pdu, const struct stat *stbuf,
uint64_t *path)
{
+ const int ino_hash_bits = qid_inode_prefix_hash_bits(pdu, stbuf->st_dev);
QppEntry lookup = {
.dev = stbuf->st_dev,
- .ino_prefix = (uint16_t) (stbuf->st_ino >> 48)
+ .ino_prefix = (uint16_t) (stbuf->st_ino >> (64 - ino_hash_bits))
}, *val;
uint32_t hash = qpp_hash(lookup);
val = qht_lookup(&pdu->s->qpp_table, &lookup, hash);
if (!val) {
- if (pdu->s->qp_prefix_next == 0) {
- /* we ran out of prefixes */
+ if (pdu->s->qp_affix_next == 0) {
+ /* we ran out of affixes */
warn_report_once(
"9p: Potential degraded performance of inode remapping"
);
@@ -693,12 +892,13 @@ static int qid_path_prefixmap(V9fsPDU *pdu, const struct stat *stbuf,
val = g_malloc0(sizeof(QppEntry));
*val = lookup;
- /* new unique inode prefix and device combo */
- val->qp_prefix = pdu->s->qp_prefix_next++;
+ /* new unique inode affix and device combo */
+ val->qp_affix_index = pdu->s->qp_affix_next++;
+ val->qp_affix = affixForIndex(val->qp_affix_index);
qht_insert(&pdu->s->qpp_table, val, hash, NULL);
}
-
- *path = ((uint64_t)val->qp_prefix << 48) | (stbuf->st_ino & QPATH_INO_MASK);
+ /* assuming generated affix to be suffix type, not prefix */
+ *path = (stbuf->st_ino << val->qp_affix.bits) | val->qp_affix.value;
return 0;
}
@@ -709,7 +909,7 @@ static int stat_to_qid(V9fsPDU *pdu, const struct stat *stbuf, V9fsQID *qidp)
if (pdu->s->ctx.export_flags & V9FS_REMAP_INODES) {
/* map inode+device to qid path (fast path) */
- err = qid_path_prefixmap(pdu, stbuf, &qidp->path);
+ err = qid_path_suffixmap(pdu, stbuf, &qidp->path);
if (err == -ENFILE) {
/* fast path didn't work, fall back to full map */
err = qid_path_fullmap(pdu, stbuf, &qidp->path);
@@ -3876,8 +4076,15 @@ int v9fs_device_realize_common(V9fsState *s, const V9fsTransport *t,
s->dev_id = stat.st_dev;
+ /* init inode remapping : */
+ /* hash table for variable length inode suffixes */
+ qpd_table_init(&s->qpd_table);
+ /* hash table for slow/full inode remapping (most users won't need it) */
+ qpf_table_init(&s->qpf_table);
+ /* hash table for quick inode remapping */
qpp_table_init(&s->qpp_table);
- s->qp_prefix_next = 1; /* reserve 0 to detect overflow */
+ s->qp_ndevices = 0;
+ s->qp_affix_next = 1; /* reserve 0 to detect overflow */
s->qp_fullpath_next = 1;
s->ctx.fst = &fse->fst;
@@ -3901,6 +4108,7 @@ void v9fs_device_unrealize_common(V9fsState *s, Error **errp)
fsdev_throttle_cleanup(s->ctx.fst);
}
g_free(s->tag);
+ qp_table_destroy(&s->qpd_table);
qp_table_destroy(&s->qpp_table);
qp_table_destroy(&s->qpf_table);
g_free(s->ctx.fs_root);
diff --git a/hw/9pfs/9p.h b/hw/9pfs/9p.h
index 35a362c0d797..3904f8290131 100644
--- a/hw/9pfs/9p.h
+++ b/hw/9pfs/9p.h
@@ -236,13 +236,49 @@ struct V9fsFidState
V9fsFidState *rclm_lst;
};
-#define QPATH_INO_MASK ((1ULL << 48) - 1)
+typedef enum AffixType_t {
+ AffixType_Prefix,
+ AffixType_Suffix, /* A.k.a. postfix. */
+} AffixType_t;
+
+/**
+ * @brief Unique affix of variable length.
+ *
+ * An affix is (currently) either a suffix or a prefix, which is either
+ * going to be prepended (prefix) or appended (suffix) with some other
+ * number for the goal to generate unique numbers. Accordingly the
+ * suffixes (or prefixes) we generate @b must all have the mathematical
+ * property of being suffix-free (or prefix-free in case of prefixes)
+ * so that no matter what number we concatenate the affix with, that we
+ * always reliably get unique numbers as result after concatenation.
+ */
+typedef struct VariLenAffix {
+ AffixType_t type; /* Whether this affix is a suffix or a prefix. */
+ uint64_t value; /* Actual numerical value of this affix. */
+ /*
+ * Lenght of the affix, that is how many (of the lowest) bits of @c value
+ * must be used for appending/prepending this affix to its final resulting,
+ * unique number.
+ */
+ int bits;
+} VariLenAffix;
+
+/* See qid_inode_prefix_hash_bits(). */
+typedef struct {
+ dev_t dev; /* FS device on host. */
+ /*
+ * How many (high) bits of the original inode number shall be used for
+ * hashing.
+ */
+ int prefix_bits;
+} QpdEntry;
/* QID path prefix entry, see stat_to_qid */
typedef struct {
dev_t dev;
uint16_t ino_prefix;
- uint16_t qp_prefix;
+ uint32_t qp_affix_index;
+ VariLenAffix qp_affix;
} QppEntry;
/* QID path full entry, as above */
@@ -274,9 +310,11 @@ struct V9fsState
V9fsConf fsconf;
V9fsQID root_qid;
dev_t dev_id;
+ struct qht qpd_table;
struct qht qpp_table;
struct qht qpf_table;
- uint16_t qp_prefix_next;
+ uint64_t qp_ndevices; /* Amount of entries in qpd_table. */
+ uint16_t qp_affix_next;
uint64_t qp_fullpath_next;
};
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PULL v2 8/8] MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes"
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
` (6 preceding siblings ...)
2019-10-10 13:18 ` [PULL v2 7/8] 9p: Use variable length suffixes for inode remapping Greg Kurz
@ 2019-10-10 13:18 ` Greg Kurz
2019-10-14 14:08 ` [PULL v2 0/8] 9p patches 2019-10-10 Peter Maydell
8 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-10 13:18 UTC (permalink / raw)
To: qemu-devel; +Cc: Peter Maydell, Greg Kurz
Latest submissions for 9pfs made me realize that I no longer have time
and motivation to actively support it. I'll stay around for odd fixes
though.
Signed-off-by: Greg Kurz <groug@kaod.org>
---
MAINTAINERS | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index 3ca814850e0c..fe4dc51b081f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1517,7 +1517,7 @@ F: tests/virtio-balloon-test.c
virtio-9p
M: Greg Kurz <groug@kaod.org>
-S: Supported
+S: Odd Fixes
F: hw/9pfs/
X: hw/9pfs/xen-9p*
F: fsdev/
--
2.21.0
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PULL v2 0/8] 9p patches 2019-10-10
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
` (7 preceding siblings ...)
2019-10-10 13:18 ` [PULL v2 8/8] MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes" Greg Kurz
@ 2019-10-14 14:08 ` Peter Maydell
2019-10-14 15:46 ` Greg Kurz
8 siblings, 1 reply; 11+ messages in thread
From: Peter Maydell @ 2019-10-14 14:08 UTC (permalink / raw)
To: Greg Kurz; +Cc: QEMU Developers
On Thu, 10 Oct 2019 at 14:18, Greg Kurz <groug@kaod.org> wrote:
>
> The following changes since commit 98b2e3c9ab3abfe476a2b02f8f51813edb90e72d:
>
> Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging (2019-10-08 16:08:35 +0100)
>
> are available in the Git repository at:
>
> https://github.com/gkurz/qemu.git tags/9p-next-2019-10-10
>
> for you to fetch changes up to e410bbc52a7a246858af54ce1842fadf26f39ff7:
>
> MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes" (2019-10-10 12:36:28 +0200)
>
> ----------------------------------------------------------------
> The most notable change is that we now detect cross-device setups in the
> host since it may cause inode number collision and mayhem in the guest.
> A new fsdev property is added for the user to choose the appropriate
> policy to handle that: either remap all inode numbers or fail I/Os to
> another host device or just print out a warning (default behaviour).
>
> This is also my last PR as _active_ maintainer of 9pfs.
>
> ----------------------------------------------------------------
Applied, thanks.
Please update the changelog at https://wiki.qemu.org/ChangeLog/4.2
for any user-visible changes.
Thanks for all your work on the 9pfs code!
-- PMM
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PULL v2 0/8] 9p patches 2019-10-10
2019-10-14 14:08 ` [PULL v2 0/8] 9p patches 2019-10-10 Peter Maydell
@ 2019-10-14 15:46 ` Greg Kurz
0 siblings, 0 replies; 11+ messages in thread
From: Greg Kurz @ 2019-10-14 15:46 UTC (permalink / raw)
To: Peter Maydell; +Cc: QEMU Developers
On Mon, 14 Oct 2019 15:08:52 +0100
Peter Maydell <peter.maydell@linaro.org> wrote:
> On Thu, 10 Oct 2019 at 14:18, Greg Kurz <groug@kaod.org> wrote:
> >
> > The following changes since commit 98b2e3c9ab3abfe476a2b02f8f51813edb90e72d:
> >
> > Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging (2019-10-08 16:08:35 +0100)
> >
> > are available in the Git repository at:
> >
> > https://github.com/gkurz/qemu.git tags/9p-next-2019-10-10
> >
> > for you to fetch changes up to e410bbc52a7a246858af54ce1842fadf26f39ff7:
> >
> > MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes" (2019-10-10 12:36:28 +0200)
> >
> > ----------------------------------------------------------------
> > The most notable change is that we now detect cross-device setups in the
> > host since it may cause inode number collision and mayhem in the guest.
> > A new fsdev property is added for the user to choose the appropriate
> > policy to handle that: either remap all inode numbers or fail I/Os to
> > another host device or just print out a warning (default behaviour).
> >
> > This is also my last PR as _active_ maintainer of 9pfs.
> >
> > ----------------------------------------------------------------
>
>
> Applied, thanks.
>
> Please update the changelog at https://wiki.qemu.org/ChangeLog/4.2
> for any user-visible changes.
>
> Thanks for all your work on the 9pfs code!
>
Thanks Peter, and many others, for your valuable help.
As now indicated in MAINTAINERS, I'll stay around for odd fixes, one
of which that comes to mind is to deprecate the "proxy" backend and
its companion virtfs-proxy-helper binary at some point since it isn't
supported by libvirt and will be superseded soon by "virtio-fs".
Cheers,
--
Greg
> -- PMM
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2019-10-14 15:49 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-10-10 13:18 [PULL v2 0/8] 9p patches 2019-10-10 Greg Kurz
2019-10-10 13:18 ` [PULL v2 1/8] 9p: unsigned type for type, version, path Greg Kurz
2019-10-10 13:18 ` [PULL v2 2/8] 9p: Simplify error path of v9fs_device_realize_common() Greg Kurz
2019-10-10 13:18 ` [PULL v2 3/8] fsdev: Add return value to fsdev_throttle_parse_opts() Greg Kurz
2019-10-10 13:18 ` [PULL v2 4/8] 9p: Treat multiple devices on one export as an error Greg Kurz
2019-10-10 13:18 ` [PULL v2 5/8] 9p: Added virtfs option 'multidevs=remap|forbid|warn' Greg Kurz
2019-10-10 13:18 ` [PULL v2 6/8] 9p: stat_to_qid: implement slow path Greg Kurz
2019-10-10 13:18 ` [PULL v2 7/8] 9p: Use variable length suffixes for inode remapping Greg Kurz
2019-10-10 13:18 ` [PULL v2 8/8] MAINTAINERS: Downgrade status of virtio-9p to "Odd Fixes" Greg Kurz
2019-10-14 14:08 ` [PULL v2 0/8] 9p patches 2019-10-10 Peter Maydell
2019-10-14 15:46 ` Greg Kurz
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).