vdpa: Fix file descriptor leak on get features error
PATCH 1/1 vdpa: Fix file descriptor leak on get features error
File descriptor vdpa_device_fd is not free in the case of returning error from vhost_vdpa_get_features. Fixing it by making all errors go to the same error path.
Resolves: Coverity CID 1490785
--- net/vhost-vdpa.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
net/vhost-vdpa.c +2 -2
@@ -566,7 +566,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name, g_autofree NetClientState **ncs = NULL;g_autoptr (VhostIOVATree) iova_tree =NULL ;NetClientState *nc; -int queue_pairs, r, i, has_cvq =0 ; +int queue_pairs, r, i =0 , has_cvq =0 ;assert (netdev->type ==NET_CLIENT_DRIVER_VHOST_VDPA ); opts = &netdev->u.vhost_vdpa;
@@ -582,7 +582,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name, r =vhost_vdpa_get_features (vdpa_device_fd, &features, errp);if (unlikely (r <0 )) { -return r; + goto err; } queue_pairs = vhost_vdpa_get_max_queue_pairs(vdpa_device_fd, features,
Signature · 1 line
2.31.1
Quoted 26 lines · Eugenio Pérez
File descriptor
vdpa_device_fdis not free in the case of returning error fromvhost_vdpa_get_features. Fixing it by making all errors go to the same error path.Resolves: Coverity CID 1490785
Fixes 8170ab3f43 ("vdpa: Extract get features part from vhost_vdpa_get_max_queue_pairs")Signed-off-by Eugenio Pérez <[email protected]>--- net/vhost-vdpa.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)net/vhost-vdpa.c +1
-1@@ -566,7 +566,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name, g_autofree NetClientState **ncs = NULL;g_autoptr (VhostIOVATree) iova_tree =NULL ;NetClientState *nc; -int queue_pairs, r, i, has_cvq =0 ; +int queue_pairs, r, i =0 , has_cvq =0 ;assert (netdev->type ==NET_CLIENT_DRIVER_VHOST_VDPA ); opts = &netdev->u.vhost_vdpa;
net/vhost-vdpa.c int net_init_vhost_vdpa(const Netdev *netdev, const char *name, · line 585
Quoted 5 lines
net/vhost-vdpa.c +1
-1@@ -582,7 +582,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name, r =vhost_vdpa_get_features (vdpa_device_fd, &features, errp);if (unlikely (r <0 )) { -return r; + goto err;
Why don't you use "goto err_svq"?
Thanks,
Laurent
Quoted 29 lines · Laurent Vivier
Quoted 27 lines · Eugenio Pérez
File descriptor
vdpa_device_fdis not free in the case of returning error fromvhost_vdpa_get_features. Fixing it by making all errors go to the same error path.Resolves: Coverity CID 1490785
Fixes 8170ab3f43 ("vdpa: Extract get features part from vhost_vdpa_get_max_queue_pairs")Signed-off-by Eugenio Pérez <[email protected]>--- net/vhost-vdpa.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)net/vhost-vdpa.c +1
-1@@ -566,7 +566,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name, g_autofree NetClientState **ncs = NULL;g_autoptr (VhostIOVATree) iova_tree =NULL ;NetClientState *nc; -int queue_pairs, r, i, has_cvq =0 ; +int queue_pairs, r, i =0 , has_cvq =0 ;assert (netdev->type ==NET_CLIENT_DRIVER_VHOST_VDPA ); opts = &netdev->u.vhost_vdpa;@@ -582,7 +582,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,
Quoted 6 lines
net/vhost-vdpa.c int net_init_vhost_vdpa(const Netdev *netdev, const char *name, · line 585
Quoted 4 lines
net/vhost-vdpa.c +1
-1r =vhost_vdpa_get_features (vdpa_device_fd, &features, errp);if (unlikely (r <0 )) { -return r; + goto err;Why don't you use "goto
err_svq"?
This patch is kind of half of the idea discussed in the thread where this problem was reported, which is that by setting i = 0 you can then consistently have all the error handling be 'goto err' and that frees everything that needs to be freed regardless of whether it's called after or before the initialization of the ncs[] entries. But it doesn't do the other half of the job, which is making all the other error handling code in the function also use 'goto err', so it looks a bit odd as it stands.
-- PMM
Quoted 38 lines · Peter Maydell
Quoted 29 lines · Laurent Vivier
Quoted 27 lines · Eugenio Pérez
File descriptor
vdpa_device_fdis not free in the case of returning error fromvhost_vdpa_get_features. Fixing it by making all errors go to the same error path.Resolves: Coverity CID 1490785
Fixes 8170ab3f43 ("vdpa: Extract get features part from vhost_vdpa_get_max_queue_pairs")Signed-off-by Eugenio Pérez <[email protected]>--- net/vhost-vdpa.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)net/vhost-vdpa.c +1
-1@@ -566,7 +566,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name, g_autofree NetClientState **ncs = NULL;g_autoptr (VhostIOVATree) iova_tree =NULL ;NetClientState *nc; -int queue_pairs, r, i, has_cvq =0 ; +int queue_pairs, r, i =0 , has_cvq =0 ;assert (netdev->type ==NET_CLIENT_DRIVER_VHOST_VDPA ); opts = &netdev->u.vhost_vdpa;@@ -582,7 +582,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,Quoted 6 lines
net/vhost-vdpa.c int net_init_vhost_vdpa(const Netdev *netdev, const char *name, · line 585
Quoted 4 lines
net/vhost-vdpa.c +1
-1r =vhost_vdpa_get_features (vdpa_device_fd, &features, errp);if (unlikely (r <0 )) { -return r; + goto err;Why don't you use "goto
err_svq"?
Quoted 9 lines
This patch is kind of half of the idea discussed in the thread where this problem was reported, which is that by setting i = 0 you can then consistently have all the error handling be 'goto err' and that frees everything that needs to be freed regardless of whether it's called after or before the initialization of the ncs[] entries. But it doesn't do the other half of the job, which is making all the other error handling code in the function also use 'goto err', so it looks a bit odd as it stands.
That's right, I thought just fixing the issue about the leaked file descriptor was the right thing to do in the hard feature freeze, and that other part should be left for the next dev phase. Is the unified error handling code acceptable for this period? I can send a patch either on top of this one or squashed if so for sure.
Thanks!
Quoted 49 lines · Eugenio Perez Martin
Quoted 38 lines · Peter Maydell
Quoted 29 lines · Laurent Vivier
Quoted 27 lines · Eugenio Pérez
File descriptor
vdpa_device_fdis not free in the case of returning error fromvhost_vdpa_get_features. Fixing it by making all errors go to the same error path.Resolves: Coverity CID 1490785
Fixes 8170ab3f43 ("vdpa: Extract get features part from vhost_vdpa_get_max_queue_pairs")Signed-off-by Eugenio Pérez <[email protected]>--- net/vhost-vdpa.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)net/vhost-vdpa.c +1
-1@@ -566,7 +566,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name, g_autofree NetClientState **ncs = NULL;g_autoptr (VhostIOVATree) iova_tree =NULL ;NetClientState *nc; -int queue_pairs, r, i, has_cvq =0 ; +int queue_pairs, r, i =0 , has_cvq =0 ;assert (netdev->type ==NET_CLIENT_DRIVER_VHOST_VDPA ); opts = &netdev->u.vhost_vdpa;@@ -582,7 +582,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,Quoted 6 lines
net/vhost-vdpa.c int net_init_vhost_vdpa(const Netdev *netdev, const char *name, · line 585
Quoted 4 lines
net/vhost-vdpa.c +1
-1r =vhost_vdpa_get_features (vdpa_device_fd, &features, errp);if (unlikely (r <0 )) { -return r; + goto err;Why don't you use "goto
err_svq"?Quoted 9 lines
This patch is kind of half of the idea discussed in the thread where this problem was reported, which is that by setting i = 0 you can then consistently have all the error handling be 'goto err' and that frees everything that needs to be freed regardless of whether it's called after or before the initialization of the ncs[] entries. But it doesn't do the other half of the job, which is making all the other error handling code in the function also use 'goto err', so it looks a bit odd as it stands.
Quoted 7 lines
That's right, I thought just fixing the issue about the leaked file descriptor was the right thing to do in the hard feature freeze, and that other part should be left for the next dev phase. Is the unified error handling code acceptable for this period? I can send a patch either on top of this one or squashed if so for sure.
Thanks!
Yea I'd prefer a minimal patch, unless others object strongly.
Signature · 1 line
MST