As Miklos reported and suggested:
This pattern repeats two times in trace_uprobe.c and in
kernel/events/core.c as well:
ret = kern_path(filename, LOOKUP_FOLLOW, &path);
if (ret)
goto fail_address_parse;
inode = igrab(d_inode(path.dentry));
path_put(&path);
And it's wrong. You can only hold a reference to the inode if you
have an active ref to the superblock as well (which is normally
through path.mnt) or holding s_umount.
This way unmounting the containing filesystem while the tracepoint is
active will give you the "VFS: Busy inodes after unmount..." message
and a crash when the inode is finally put.
Solution: store path instead of inode.
This patch fixes two instances in trace_uprobe.c. struct path is added to
struct trace_uprobe to keep the inode and containing mount point
referenced.
Fixes: f3f096cfedf8 ("tracing: Provide trace events interface for uprobes")
Fixes: 33ea4b24277b ("perf/core: Implement the 'perf_uprobe' PMU")
Cc: Steven Rostedt <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Howard McLauchlan <[email protected]>
Cc: Josef Bacik <[email protected]>
Cc: Srikar Dronamraju <[email protected]>
Cc: Miklos Szeredi <[email protected]>
Reported-by: Miklos Szeredi <[email protected]>
Signed-off-by: Song Liu <[email protected]>
---
kernel/trace/trace_uprobe.c | 35 ++++++++++++++---------------------
1 file changed, 14 insertions(+), 21 deletions(-)
diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
index 34fd0e0..ac89287 100644
--- a/kernel/trace/trace_uprobe.c
+++ b/kernel/trace/trace_uprobe.c
@@ -55,6 +55,7 @@ struct trace_uprobe {
struct list_head list;
struct trace_uprobe_filter filter;
struct uprobe_consumer consumer;
+ struct path path;
struct inode *inode;
char *filename;
unsigned long offset;
@@ -289,7 +290,7 @@ static void free_trace_uprobe(struct trace_uprobe *tu)
for (i = 0; i < tu->tp.nr_args; i++)
traceprobe_free_probe_arg(&tu->tp.args[i]);
- iput(tu->inode);
+ path_put(&tu->path);
kfree(tu->tp.call.class->system);
kfree(tu->tp.call.name);
kfree(tu->filename);
@@ -363,7 +364,6 @@ static int register_trace_uprobe(struct trace_uprobe *tu)
static int create_trace_uprobe(int argc, char **argv)
{
struct trace_uprobe *tu;
- struct inode *inode;
char *arg, *event, *group, *filename;
char buf[MAX_EVENT_NAME_LEN];
struct path path;
@@ -371,7 +371,6 @@ static int create_trace_uprobe(int argc, char **argv)
bool is_delete, is_return;
int i, ret;
- inode = NULL;
ret = 0;
is_delete = false;
is_return = false;
@@ -437,21 +436,16 @@ static int create_trace_uprobe(int argc, char **argv)
}
/* Find the last occurrence, in case the path contains ':' too. */
arg = strrchr(argv[1], ':');
- if (!arg) {
- ret = -EINVAL;
- goto fail_address_parse;
- }
+ if (!arg)
+ return -EINVAL;
*arg++ = '\0';
filename = argv[1];
ret = kern_path(filename, LOOKUP_FOLLOW, &path);
if (ret)
- goto fail_address_parse;
-
- inode = igrab(d_real_inode(path.dentry));
- path_put(&path);
+ return ret;
- if (!inode || !S_ISREG(inode->i_mode)) {
+ if (!d_is_reg(path.dentry)) {
ret = -EINVAL;
goto fail_address_parse;
}
@@ -490,7 +484,7 @@ static int create_trace_uprobe(int argc, char **argv)
goto fail_address_parse;
}
tu->offset = offset;
- tu->inode = inode;
+ tu->path = path;
tu->filename = kstrdup(filename, GFP_KERNEL);
if (!tu->filename) {
@@ -558,7 +552,7 @@ static int create_trace_uprobe(int argc, char **argv)
return ret;
fail_address_parse:
- iput(inode);
+ path_put(&path);
pr_info("Failed to parse address or file.\n");
@@ -922,6 +916,7 @@ probe_event_enable(struct trace_uprobe *tu, struct trace_event_file *file,
goto err_flags;
tu->consumer.filter = filter;
+ tu->inode = d_real_inode(tu->path.dentry);
ret = uprobe_register(tu->inode, tu->offset, &tu->consumer);
if (ret)
goto err_buffer;
@@ -967,6 +962,7 @@ probe_event_disable(struct trace_uprobe *tu, struct trace_event_file *file)
WARN_ON(!uprobe_filter_is_empty(&tu->filter));
uprobe_unregister(tu->inode, tu->offset, &tu->consumer);
+ tu->inode = NULL;
tu->tp.flags &= file ? ~TP_FLAG_TRACE : ~TP_FLAG_PROFILE;
uprobe_buffer_disable();
@@ -1337,7 +1333,6 @@ struct trace_event_call *
create_local_trace_uprobe(char *name, unsigned long offs, bool is_return)
{
struct trace_uprobe *tu;
- struct inode *inode;
struct path path;
int ret;
@@ -1345,11 +1340,8 @@ create_local_trace_uprobe(char *name, unsigned long offs, bool is_return)
if (ret)
return ERR_PTR(ret);
- inode = igrab(d_inode(path.dentry));
- path_put(&path);
-
- if (!inode || !S_ISREG(inode->i_mode)) {
- iput(inode);
+ if (!d_is_reg(path.dentry)) {
+ path_put(&path);
return ERR_PTR(-EINVAL);
}
@@ -1364,11 +1356,12 @@ create_local_trace_uprobe(char *name, unsigned long offs, bool is_return)
if (IS_ERR(tu)) {
pr_info("Failed to allocate trace_uprobe.(%d)\n",
(int)PTR_ERR(tu));
+ path_put(&path);
return ERR_CAST(tu);
}
tu->offset = offs;
- tu->inode = inode;
+ tu->path = path;
tu->filename = kstrdup(name, GFP_KERNEL);
init_trace_event_call(tu, &tu->tp.call);
--
2.9.5
Caller of uprobe_register is required to keep the inode and containing
mount point referenced.
There was misuse of igrab() in uprobes.c and trace_uprobe.c. This is
because igrab() will not prevent umount of the containing mount point.
To fix this, we added path to struct trace_uprobe, which keeps the inode
and containing mount reference.
For uprobes.c, it is not necessary to call igrab() in uprobe_register(),
as the caller is required to keep the inode reference. The igrab() is
removed and comments on this requirement is added to uprobe_register().
Link: http://lkml.kernel.org/r/CAELBmZB2XX=qEOLAdvGG4cPx4GEntcSnWQquJLUK1ongRj35cA@mail.gmail.com
Cc: Steven Rostedt <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Howard McLauchlan <[email protected]>
Cc: Josef Bacik <[email protected]>
Cc: Srikar Dronamraju <[email protected]>
Cc: Miklos Szeredi <[email protected]>
Signed-off-by: Song Liu <[email protected]>
---
kernel/events/uprobes.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index ce6848e..1725b90 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -491,7 +491,7 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset)
if (!uprobe)
return NULL;
- uprobe->inode = igrab(inode);
+ uprobe->inode = inode;
uprobe->offset = offset;
init_rwsem(&uprobe->register_rwsem);
init_rwsem(&uprobe->consumer_rwsem);
@@ -502,7 +502,6 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset)
if (cur_uprobe) {
kfree(uprobe);
uprobe = cur_uprobe;
- iput(inode);
}
return uprobe;
@@ -701,7 +700,6 @@ static void delete_uprobe(struct uprobe *uprobe)
rb_erase(&uprobe->rb_node, &uprobes_tree);
spin_unlock(&uprobes_treelock);
RB_CLEAR_NODE(&uprobe->rb_node); /* for uprobe_is_active() */
- iput(uprobe->inode);
put_uprobe(uprobe);
}
@@ -873,7 +871,8 @@ static void __uprobe_unregister(struct uprobe *uprobe, struct uprobe_consumer *u
* tuple). Creation refcount stops uprobe_unregister from freeing the
* @uprobe even before the register operation is complete. Creation
* refcount is released when the last @uc for the @uprobe
- * unregisters.
+ * unregisters. Caller of uprobe_register() is required to keep @inode
+ * (and the containing mount) referenced.
*
* Return errno if it cannot successully install probes
* else return 0 (success)
--
2.9.5
On Mon, Apr 23, 2018 at 7:21 PM, Song Liu <[email protected]> wrote:
> As Miklos reported and suggested:
>
> This pattern repeats two times in trace_uprobe.c and in
> kernel/events/core.c as well:
>
> ret = kern_path(filename, LOOKUP_FOLLOW, &path);
> if (ret)
> goto fail_address_parse;
>
> inode = igrab(d_inode(path.dentry));
> path_put(&path);
>
> And it's wrong. You can only hold a reference to the inode if you
> have an active ref to the superblock as well (which is normally
> through path.mnt) or holding s_umount.
>
> This way unmounting the containing filesystem while the tracepoint is
> active will give you the "VFS: Busy inodes after unmount..." message
> and a crash when the inode is finally put.
>
> Solution: store path instead of inode.
>
> This patch fixes two instances in trace_uprobe.c. struct path is added to
> struct trace_uprobe to keep the inode and containing mount point
> referenced.
>
> Fixes: f3f096cfedf8 ("tracing: Provide trace events interface for uprobes")
> Fixes: 33ea4b24277b ("perf/core: Implement the 'perf_uprobe' PMU")
> Cc: Steven Rostedt <[email protected]>
> Cc: Ingo Molnar <[email protected]>
> Cc: Howard McLauchlan <[email protected]>
> Cc: Josef Bacik <[email protected]>
> Cc: Srikar Dronamraju <[email protected]>
> Cc: Miklos Szeredi <[email protected]>
> Reported-by: Miklos Szeredi <[email protected]>
> Signed-off-by: Song Liu <[email protected]>
Acked-by: Miklos Szeredi <[email protected]>
On Mon, Apr 23, 2018 at 7:21 PM, Song Liu <[email protected]> wrote:
> Caller of uprobe_register is required to keep the inode and containing
> mount point referenced.
>
> There was misuse of igrab() in uprobes.c and trace_uprobe.c. This is
> because igrab() will not prevent umount of the containing mount point.
> To fix this, we added path to struct trace_uprobe, which keeps the inode
> and containing mount reference.
>
> For uprobes.c, it is not necessary to call igrab() in uprobe_register(),
> as the caller is required to keep the inode reference. The igrab() is
> removed and comments on this requirement is added to uprobe_register().
>
> Link: http://lkml.kernel.org/r/CAELBmZB2XX=qEOLAdvGG4cPx4GEntcSnWQquJLUK1ongRj35cA@mail.gmail.com
> Cc: Steven Rostedt <[email protected]>
> Cc: Ingo Molnar <[email protected]>
> Cc: Howard McLauchlan <[email protected]>
> Cc: Josef Bacik <[email protected]>
> Cc: Srikar Dronamraju <[email protected]>
> Cc: Miklos Szeredi <[email protected]>
> Signed-off-by: Song Liu <[email protected]>
Acked-by: Miklos Szeredi <[email protected]>