uprobes: Shift ->readpage check from __copy_insn() to uprobe_register()

PATCH In review Oleg Nesterov 3 participants 6 messages May 19, 2014 – Jun 5, 2014

Subsystems (2)

perf events uprobes

lore

Share & tools

lore

Oleg Nesterov

author maintainer

2014-05-19 18:41 UTC #0 lore raw

PATCH 1/1 uprobes: Shift ->readpage check from __copy_insn() to uprobe_register()

Sorry for double-posting, but it seems that this patch didn't reach lkml. Let me resend it just on case. Plus another patch in reply, on top of this change.

-------------------------------------------------------------------------------
Subject: [PATCH] uprobes: Shift ->readpage check from __copy_insn() to uprobe_register()

copy_insn() fails with -EIO if ->readpage == NULL, but this error is not propagated unless uprobe_register() path finds ->mm which already mmaps this file. In this case (say) "perf record" does not actually install the probe, but the user can't know about this.

Move this check into uprobe_register() so that this problem can be detected earlier and reported to user.

Note: this is still not perfect,

- copy_insn() and arch_uprobe_analyze_insn() should be called by uprobe_register() but this is not simple, we need vm_file for read_mapping_page() (although perhaps we can pass NULL), and we need ->mm for is_64bit_mm() (although this logic is broken anyway).

- uprobe_register() should be called by create_trace_uprobe(), not by probe_event_enable(), so that an error can be detected at "perf probe -x" time. This also needs more changes in the core uprobe code, uprobe register/unregister interface was poorly designed from the very beginning.

Reported-by Denys Vlasenko <[email protected]>

Signed-off-by Oleg Nesterov <[email protected]>

---
 kernel/events/uprobes.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

kernel/events/uprobes.c +3 -3

index 3b02c72..c56b13e 100644
@@ -536,9 +536,6 @@ static int __copy_insn(struct address_space *mapping, struct file *filp,
 			void *insn, int nbytes, loff_t offset)
 {
 	struct page *page;
-
-	if (!mapping->a_ops->readpage)
-		return -EIO;
 	/*
 	 * Ensure that the page that has the original instruction is
 	 * populated and in page-cache.

@@ -879,6 +876,9 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *
 	if (!uc->handler && !uc->ret_handler)
 		return -EINVAL;
 
+	/* copy_insn()->read_mapping_page() needs ->readpage() */
+	if (!inode->i_mapping->a_ops->readpage)
+		return -EIO;
 	/* Racy, just to catch the obvious mistakes */
 	if (offset > i_size_read(inode))
 		return -EINVAL;

Signature · 1 line

1.5.5.1

Oleg Nesterov

author maintainer

2014-05-19 18:42 UTC #1 lore raw

PATCH 1/1 uprobes: Teach copy_insn() to support tmpfs

tmpfs is widely used but as Denys reports shmem_aops doesn't have ->readpage() and thus you can't probe a binary on this filesystem.

As Hugh suggested we can use shmem_read_mapping_page() in this case, just we need to check shmem_mapping() if ->readpage == NULL.

Reported-by Denys Vlasenko <[email protected]>

Signed-off-by Oleg Nesterov <[email protected]>

---
 kernel/events/uprobes.c |   15 ++++++++++-----
 1 files changed, 10 insertions(+), 5 deletions(-)

kernel/events/uprobes.c +10 -5

index c56b13e..6bfb671 100644
@@ -36,6 +36,7 @@
 #include "../../mm/internal.h"	/* munlock_vma_page */
 #include <linux/percpu-rwsem.h>
 #include <linux/task_work.h>
+#include <linux/shmem_fs.h>
 
 #include <linux/uprobes.h>

@@ -537,10 +538,14 @@ static int __copy_insn(struct address_space *mapping, struct file *filp,
 {
 	struct page *page;
 	/*
-	 * Ensure that the page that has the original instruction is
-	 * populated and in page-cache.
+	 * Ensure that the page that has the original instruction is populated
+	 * and in page-cache. If ->readpage == NULL it must be shmem_mapping(),
+	 * see uprobe_register().
 	 */
-	page = read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT, filp);
+	if (mapping->a_ops->readpage)
+		page = read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT, filp);
+	else
+		page = shmem_read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT);
 	if (IS_ERR(page))
 		return PTR_ERR(page);

@@ -876,8 +881,8 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *
 	if (!uc->handler && !uc->ret_handler)
 		return -EINVAL;
 
-	/* copy_insn()->read_mapping_page() needs ->readpage() */
-	if (!inode->i_mapping->a_ops->readpage)
+	/* copy_insn() uses read_mapping_page() or shmem_read_mapping_page() */
+	if (!inode->i_mapping->a_ops->readpage && !shmem_mapping(inode->i_mapping))
 		return -EIO;
 	/* Racy, just to catch the obvious mistakes */
 	if (offset > i_size_read(inode))

Signature · 1 line

1.5.5.1

Srikar Dronamraju

Acked-by on v1 1/1

2014-05-21 08:54 UTC #2 lore raw

Re: [PATCH] uprobes: Teach copy_insn() to support tmpfs

Quoted 9 lines · Oleg Nesterov

tmpfs is widely used but as Denys reports shmem_aops doesn't have ->readpage() and thus you can't probe a binary on this filesystem.
As Hugh suggested we can use shmem_read_mapping_page() in this case, just we need to check shmem_mapping() if ->readpage == NULL.
Reported-by Denys Vlasenko <[email protected]>
Suggested-by Hugh Dickins <[email protected]>
Signed-off-by Oleg Nesterov <[email protected]>

Acked-by Srikar Dronamraju <[email protected]>

Signature · 2 lines

Thanks and Regards
Srikar Dronamraju

tip-bot for Oleg Nesterov

merge bot

2014-06-05 14:40 UTC #3 lore raw

[tip:perf/core] uprobes: Teach copy_insn() to support tmpfs

In reply to Oleg Nesterov

uprobes: Teach copy_insn() to support tmpfs

Commit-ID:  40814f6805a524130a5ec313871147f7aa9b5318
Gitweb:     http://git.kernel.org/tip/40814f6805a524130a5ec313871147f7aa9b5318
Author:     Oleg Nesterov <[email protected]>
AuthorDate: Mon, 19 May 2014 20:41:36 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 5 Jun 2014 12:30:11 +0200

uprobes: Teach copy_insn() to support tmpfs

tmpfs is widely used but as Denys reports shmem_aops doesn't have ->readpage() and thus you can't probe a binary on this filesystem.

As Hugh suggested we can use shmem_read_mapping_page() in this case, just we need to check shmem_mapping() if ->readpage == NULL.

Reported-by Denys Vlasenko <[email protected]>

Signed-off-by Oleg Nesterov <[email protected]>

Acked-by Srikar Dronamraju <[email protected]>

Signed-off-by Peter Zijlstra <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc Linus Torvalds <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/uprobes.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

kernel/events/uprobes.c +10 -5

index c56b13e..6bfb671 100644
@@ -36,6 +36,7 @@
 #include "../../mm/internal.h"	/* munlock_vma_page */
 #include <linux/percpu-rwsem.h>
 #include <linux/task_work.h>
+#include <linux/shmem_fs.h>
 
 #include <linux/uprobes.h>

@@ -537,10 +538,14 @@ static int __copy_insn(struct address_space *mapping, struct file *filp,
 {
 	struct page *page;
 	/*
-	 * Ensure that the page that has the original instruction is
-	 * populated and in page-cache.
+	 * Ensure that the page that has the original instruction is populated
+	 * and in page-cache. If ->readpage == NULL it must be shmem_mapping(),
+	 * see uprobe_register().
 	 */
-	page = read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT, filp);
+	if (mapping->a_ops->readpage)
+		page = read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT, filp);
+	else
+		page = shmem_read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT);
 	if (IS_ERR(page))
 		return PTR_ERR(page);

@@ -876,8 +881,8 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *
 	if (!uc->handler && !uc->ret_handler)
 		return -EINVAL;
 
-	/* copy_insn()->read_mapping_page() needs ->readpage() */
-	if (!inode->i_mapping->a_ops->readpage)
+	/* copy_insn() uses read_mapping_page() or shmem_read_mapping_page() */
+	if (!inode->i_mapping->a_ops->readpage && !shmem_mapping(inode->i_mapping))
 		return -EIO;
 	/* Racy, just to catch the obvious mistakes */
 	if (offset > i_size_read(inode))

Srikar Dronamraju

Acked-by on v1 1/1

2014-05-21 08:53 UTC #4 lore raw

In reply to Oleg Nesterov

Quoted 23 lines · Oleg Nesterov

Sorry for double-posting, but it seems that this patch didn't reach lkml. Let me resend it just on case. Plus another patch in reply, on top of this change.
-------------------------------------------------------------------------------
Subject: [PATCH] uprobes: Shift ->readpage check from __copy_insn() to uprobe_register()
copy_insn() fails with -EIO if ->readpage == NULL, but this error is not propagated unless uprobe_register() path finds ->mm which already mmaps this file. In this case (say) "perf record" does not actually install the probe, but the user can't know about this.
Move this check into uprobe_register() so that this problem can be detected earlier and reported to user.
Note: this is still not perfect,
- copy_insn() and arch_uprobe_analyze_insn() should be called by uprobe_register() but this is not simple, we need vm_file for read_mapping_page() (although perhaps we can pass NULL), and we need ->mm for is_64bit_mm() (although this logic is broken anyway).

Quoted 8 lines

- uprobe_register() should be called by create_trace_uprobe(), not by probe_event_enable(), so that an error can be detected at "perf probe -x" time. This also needs more changes in the core uprobe code, uprobe register/unregister interface was poorly designed from the very beginning.
Reported-by Denys Vlasenko <[email protected]>
Signed-off-by Oleg Nesterov <[email protected]>

Acked-by Srikar Dronamraju <[email protected]>

Signature · 2 lines

Thanks and Regards
Srikar Dronamraju

tip-bot for Oleg Nesterov

merge bot

2014-06-05 14:39 UTC #5 lore raw

[tip:perf/core] uprobes: Shift ->readpage check from __copy_insn( ) to uprobe_register()

In reply to Oleg Nesterov

uprobes: Shift ->readpage check from __copy_insn() to uprobe_register()

Commit-ID:  41ccba029e9492dd0fc1bf5c23b72c6322a6dfe9
Gitweb:     http://git.kernel.org/tip/41ccba029e9492dd0fc1bf5c23b72c6322a6dfe9
Author:     Oleg Nesterov <[email protected]>
AuthorDate: Mon, 19 May 2014 20:40:54 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 5 Jun 2014 12:30:07 +0200

uprobes: Shift ->readpage check from __copy_insn() to uprobe_register()

Move this check into uprobe_register() so that this problem can be detected earlier and reported to user.

Note: this is still not perfect,

Reported-by Denys Vlasenko <[email protected]>

Signed-off-by Oleg Nesterov <[email protected]>

Acked-by Srikar Dronamraju <[email protected]>

Signed-off-by Peter Zijlstra <[email protected]>

Cc Hugh Dickins <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc Linus Torvalds <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/uprobes.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

kernel/events/uprobes.c +3 -3

index 3b02c72..c56b13e 100644
@@ -536,9 +536,6 @@ static int __copy_insn(struct address_space *mapping, struct file *filp,
 			void *insn, int nbytes, loff_t offset)
 {
 	struct page *page;
-
-	if (!mapping->a_ops->readpage)
-		return -EIO;
 	/*
 	 * Ensure that the page that has the original instruction is
 	 * populated and in page-cache.

@@ -879,6 +876,9 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *
 	if (!uc->handler && !uc->ret_handler)
 		return -EINVAL;
 
+	/* copy_insn()->read_mapping_page() needs ->readpage() */
+	if (!inode->i_mapping->a_ops->readpage)
+		return -EIO;
 	/* Racy, just to catch the obvious mistakes */
 	if (offset > i_size_read(inode))
 		return -EINVAL;