dma-buf: use spinlock to protect set/get name operation
Subsystems (2)
PATCH 1/1 dma-buf: use spinlock to protect set/get name operation
We introduced setname ioctl in commit bb2bb9030425 ("dma-buf:
add DMA_BUF_SET_NAME ioctls") that provides userpsace to attach a free-form name for tracking and counting shared buffers. However the d_dname callback could be called in atomic context. This call path comes from selinux that verifies all inherited open files from exec call. To verify all inherited open files, kernel would iterate all fds which need to hold spin_lock to get denty name by calling d_dname operation. In dma-buf d_dname callback, we use mutex lock to prevent the race from setname causing this issue.
This commit adds a spinlock to protect set/get name operation to fix this issue.
[ 165.617090] Call trace: [ 165.620504] ___might_sleep+0x114/0x118 [ 165.625344] __might_sleep+0x50/0x84 [ 165.629928] __mutex_lock_common+0x5c/0x10b0 [ 165.635215] mutex_lock_nested+0x40/0x50 [ 165.640157] dmabuffs_dname+0x48/0xdc [ 165.644821] d_path+0x78/0x1e4 [ 165.648870] audit_log_d_path+0x68/0x134 [ 165.653807] common_lsm_audit+0x33c/0x6f4 [ 165.658832] slow_avc_audit+0xb4/0xf0 [ 165.663503] avc_has_perm+0xdc/0x1a4 [ 165.668081] file_has_perm+0x70/0x154 [ 165.672750] match_file+0x54/0x6c [ 165.677064] iterate_fd+0x74/0xac [ 165.681369] selinux_bprm_committing_creds+0xfc/0x210 [ 165.687459] security_bprm_committing_creds+0x2c/0x40 [ 165.693546] install_exec_creds+0x1c/0x68 [ 165.698569] load_elf_binary+0x3a0/0x13c8 [ 165.703590] search_binary_handler+0xb8/0x1e4 [ 165.708964] __do_execve_file+0x6e4/0x9c8 [ 165.713984] __arm64_sys_execve+0x44/0x54 [ 165.719008] el0_svc_common+0xa8/0x168 [ 165.723765] el0_svc_handler+0x78/0x94 [ 165.728522] el0_svc+0x8/0xc
--- drivers/dma-buf/dma-buf.c | 11 +++++++---- include/linux/dma-buf.h | 2 ++ 2 files changed, 9 insertions(+), 4 deletions(-)
drivers/dma-buf/dma-buf.c +7 -4
@@ -45,10 +45,10 @@ static char *dmabuffs_dname(struct dentry *dentry, char *buffer, int buflen)size_t ret =0 ; dmabuf = dentry->d_fsdata; -dma_resv_lock (dmabuf->resv,NULL ); +spin_lock (&dmabuf->name_lock);if (dmabuf->name) ret =strlcpy (name, dmabuf->name,DMA_BUF_NAME_LEN ); -dma_resv_unlock (dmabuf->resv); +spin_unlock (&dmabuf->name_lock);return dynamic_dname (dentry, buffer, buflen,"/%s:%s" , dentry->d_name.name, ret >0 ? name :"" );
@@ -335,6 +335,7 @@ static long dma_buf_set_name(struct dma_buf *dmabuf, const char __user *buf)return PTR_ERR (name);dma_resv_lock (dmabuf->resv,NULL ); +spin_lock (&dmabuf->name_lock);if (!list_empty (&dmabuf->attachments)) { ret = -EBUSY ;kfree (name);
@@ -344,6 +345,7 @@ static long dma_buf_set_name(struct dma_buf *dmabuf, const char __user *buf) dmabuf->name = name; out_unlock: +spin_unlock (&dmabuf->name_lock);dma_resv_unlock (dmabuf->resv);return ret; }
@@ -403,10 +405,10 @@ static void dma_buf_show_fdinfo(struct seq_file *m, struct file *file)/* Don't count the temporary reference taken inside procfs seq_show */ seq_printf (m,"count:\t%ld\n" ,file_count (dmabuf->file) -1 );seq_printf (m,"exp_name:\t%s\n" , dmabuf->exp_name); -dma_resv_lock (dmabuf->resv,NULL ); +spin_lock (&dmabuf->name_lock);if (dmabuf->name)seq_printf (m,"name:\t%s\n" , dmabuf->name); -dma_resv_unlock (dmabuf->resv); +spin_unlock (&dmabuf->name_lock); }static const struct file_operations dma_buf_fops = {
@@ -561,6 +563,7 @@ struct dma_buf *dma_buf_export(const struct dma_buf_export_info *exp_info) dmabuf->file = file;mutex_init (&dmabuf->lock); +spin_lock_init (&dmabuf->name_lock);INIT_LIST_HEAD (&dmabuf->attachments);mutex_lock (&db_list.lock);
include/linux/dma-buf.h +2 -0
@@ -292,6 +292,7 @@ struct dma_buf_ops {* @exp_name: name of the exporter; useful for debugging. * @name: userspace-provided name; useful for accounting and debugging, * protected by @resv. +* @name_lock: lock to protect name. * @owner: pointer to exporter module; used for refcounting when exporter is a * kernel module. * @list_node: node for dma_buf accounting and debugging.
@@ -320,6 +321,7 @@ struct dma_buf {void *vmap_ptr;const char *exp_name;const char *name; +spinlock_t name_lock;struct module *owner;struct list_head list_node;void *priv;
Signature · 1 line
2.25.0.rc1.283.g88dfdc4193-goog
A message that is not in the archive (sent off-list or not yet received): CAO_48GEbH+JM6247KUc+XeD2xAcMsMmfNHXd1R7sLkkWPgfX7Q@mail.gmail.com
Quoted 9 lines · Sumit Semwal
Hello Martin,
Thanks for your patches - they look decent.
May I please request you to run get_maintainers.pl (the patches need to be sent to a couple of other MLs too for wider review).
Best,
Sumit.
Sorry for the late reply. Sure, I will include more MLs for wider review. Thanks for the suggestion.
Quoted 132 lines · Martin Liu
Quoted 123 lines
We introduced setname ioctl in commit
bb2bb9030425("dma-buf:
addDMA_BUF_SET_NAMEioctls") that provides userpsace to attach a free-form name for tracking and counting shared buffers. However thed_dnamecallback could be called in atomic context. This call path comes from selinux that verifies all inherited open files from exec call. To verify all inherited open files, kernel would iterate all fds which need to holdspin_lockto get denty name by callingd_dnameoperation. In dma-bufd_dnamecallback, we use mutex lock to prevent the race from setname causing this issue.This commit adds a spinlock to protect set/get name operation to fix this issue.
[ 165.617090] Call trace: [ 165.620504] ___might_sleep+0x114/0x118 [ 165.625344] __might_sleep+0x50/0x84 [ 165.629928] __mutex_lock_common+0x5c/0x10b0 [ 165.635215] mutex_lock_nested+0x40/0x50 [ 165.640157] dmabuffs_dname+0x48/0xdc [ 165.644821] d_path+0x78/0x1e4 [ 165.648870] audit_log_d_path+0x68/0x134 [ 165.653807] common_lsm_audit+0x33c/0x6f4 [ 165.658832] slow_avc_audit+0xb4/0xf0 [ 165.663503] avc_has_perm+0xdc/0x1a4 [ 165.668081] file_has_perm+0x70/0x154 [ 165.672750] match_file+0x54/0x6c [ 165.677064] iterate_fd+0x74/0xac [ 165.681369] selinux_bprm_committing_creds+0xfc/0x210 [ 165.687459] security_bprm_committing_creds+0x2c/0x40 [ 165.693546] install_exec_creds+0x1c/0x68 [ 165.698569] load_elf_binary+0x3a0/0x13c8 [ 165.703590] search_binary_handler+0xb8/0x1e4 [ 165.708964] __do_execve_file+0x6e4/0x9c8 [ 165.713984] __arm64_sys_execve+0x44/0x54 [ 165.719008] el0_svc_common+0xa8/0x168 [ 165.723765] el0_svc_handler+0x78/0x94 [ 165.728522] el0_svc+0x8/0xcSigned-off-by Martin Liu <[email protected]>--- drivers/dma-buf/dma-buf.c | 11 +++++++---- include/linux/dma-buf.h | 2 ++ 2 files changed, 9 insertions(+), 4 deletions(-)drivers/dma-buf/dma-buf.c +7
-4@@ -45,10 +45,10 @@ static char *dmabuffs_dname(struct dentry *dentry, char *buffer, int buflen)size_t ret =0 ; dmabuf = dentry->d_fsdata; -dma_resv_lock (dmabuf->resv,NULL ); +spin_lock (&dmabuf->name_lock);if (dmabuf->name) ret =strlcpy (name, dmabuf->name,DMA_BUF_NAME_LEN ); -dma_resv_unlock (dmabuf->resv); +spin_unlock (&dmabuf->name_lock);return dynamic_dname (dentry, buffer, buflen,"/%s:%s" , dentry->d_name.name, ret >0 ? name :"" );@@ -335,6 +335,7 @@ static long dma_buf_set_name(struct dma_buf *dmabuf, const char __user *buf)return PTR_ERR (name);dma_resv_lock (dmabuf->resv,NULL ); +spin_lock (&dmabuf->name_lock);if (!list_empty (&dmabuf->attachments)) { ret = -EBUSY ;kfree (name);@@ -344,6 +345,7 @@ static long dma_buf_set_name(struct dma_buf *dmabuf, const char __user *buf) dmabuf->name = name; out_unlock: +spin_unlock (&dmabuf->name_lock);dma_resv_unlock (dmabuf->resv);return ret; }@@ -403,10 +405,10 @@ static void dma_buf_show_fdinfo(struct seq_file *m, struct file *file)/* Don't count the temporary reference taken inside procfs seq_show */ seq_printf (m,"count:\t%ld\n" ,file_count (dmabuf->file) -1 );seq_printf (m,"exp_name:\t%s\n" , dmabuf->exp_name); -dma_resv_lock (dmabuf->resv,NULL ); +spin_lock (&dmabuf->name_lock);if (dmabuf->name)seq_printf (m,"name:\t%s\n" , dmabuf->name); -dma_resv_unlock (dmabuf->resv); +spin_unlock (&dmabuf->name_lock); }static const struct file_operations dma_buf_fops = {@@ -561,6 +563,7 @@ struct dma_buf *dma_buf_export(const struct dma_buf_export_info *exp_info) dmabuf->file = file;mutex_init (&dmabuf->lock); +spin_lock_init (&dmabuf->name_lock);INIT_LIST_HEAD (&dmabuf->attachments);mutex_lock (&db_list.lock);include/linux/dma-buf.h +2
-1@@ -292,6 +292,7 @@ struct dma_buf_ops {* @exp_name: name of the exporter; useful for debugging. * @name: userspace-provided name; useful for accounting and debugging, * protected by @resv. +* @name_lock: lock to protect name. * @owner: pointer to exporter module; used for refcounting when exporter is a * kernel module. * @list_node: node for dma_buf accounting and debugging. @@ -320,6 +321,7 @@ struct dma_buf {void *vmap_ptr;const char *exp_name;const char *name; +spinlock_t name_lock;struct module *owner;struct list_head list_node;void *priv; --2.25.0.rc1.283.g88dfdc4193-goog
Signature · 5 lines
Thanks and regards, Sumit Semwal Linaro Consumer Group - Kernel Team Lead Linaro.org │ Open source software for ARM SoCs