Skip to contentSkip to conversation

uprobes: ensure that uprobe->offset and ->ref_ctr_offset are properly aligned

PATCH In review Oleg Nesterov 7 participants 11 messages May 4, 2020 – Jun 10, 2020 3 messages not in the archive

Subsystems (2)
lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): CAHk-=whQt69ApMkZF8b2Q2idMDgPpPETZeeOuZg59CrOO4025w@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

Oleg Nesterov

author maintainer
#3 lore raw

PATCH 1/1 uprobes: ensure that uprobe->offset and ->ref_ctr_offset are properly aligned

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Another problem is __update_ref_ctr() which was wrong from the very beginning, it can read/write outside of kmap'ed page unless "vaddr" is aligned to sizeof(short), __uprobe_register() should check this too.

---
 kernel/events/uprobes.c | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)
kernel/events/uprobes.c +12 -4
index ece7e13f6e4a..cc2095607c74 100644
@@ -867,10 +867,6 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file,
 	if (ret)
 		goto out;
 
-	/* uprobe_write_opcode() assumes we don't cross page boundary */
-	BUG_ON((uprobe->offset & ~PAGE_MASK) +
-			UPROBE_SWBP_INSN_SIZE > PAGE_SIZE);
-
 	smp_wmb(); /* pairs with the smp_rmb() in handle_swbp() */
 	set_bit(UPROBE_COPY_INSN, &uprobe->flags);
 
@@ -1166,6 +1162,15 @@ static int __uprobe_register(struct inode *inode, loff_t offset,
 	if (offset > i_size_read(inode))
 		return -EINVAL;
 
+	/*
+	 * This ensures that copy_from_page(), copy_to_page() and
+	 * __update_ref_ctr() can't cross page boundary.
+	 */
+	if (!IS_ALIGNED(offset, UPROBE_SWBP_INSN_SIZE))
+		return -EINVAL;
+	if (!IS_ALIGNED(ref_ctr_offset, sizeof(short)))
+		return -EINVAL;
+
  retry:
 	uprobe = alloc_uprobe(inode, offset, ref_ctr_offset);
 	if (!uprobe)
@@ -2014,6 +2019,9 @@ static int is_trap_at_addr(struct mm_struct *mm, unsigned long vaddr)
 	uprobe_opcode_t opcode;
 	int result;
 
+	if (WARN_ON_ONCE(!IS_ALIGNED(vaddr, UPROBE_SWBP_INSN_SIZE)))
+		return -EINVAL;
+
 	pagefault_disable();
 	result = __get_user(opcode, (uprobe_opcode_t __user *)vaddr);
 	pagefault_enable();
Signature · 1 line
2.25.1.362.g51ebf55
#4 lore raw
Quoted 3 lines · Oleg Nesterov

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

I think the idea was that the uprobe instruction is 2 bytes and instructions are always aligned to 2 bytes on s390. (we can have 2,4 or 6 bytes).

Quoted 5 lines

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Not sure if it would have been possible to try to create a uprobe on an odd address. If yes, then the new IS_ALIGNED check certainly makes this better for s390, so the patch looks sane. Adding Vasily and Sven to double check.

Quoted 55 lines

Another problem is __update_ref_ctr() which was wrong from the very beginning, it can read/write outside of kmap'ed page unless "vaddr" is aligned to sizeof(short), __uprobe_register() should check this too.

---
 kernel/events/uprobes.c | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)
kernel/events/uprobes.c +12 -4
index ece7e13f6e4a..cc2095607c74 100644
@@ -867,10 +867,6 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file,
 	if (ret)
 		goto out;
 
-	/* uprobe_write_opcode() assumes we don't cross page boundary */
-	BUG_ON((uprobe->offset & ~PAGE_MASK) +
-			UPROBE_SWBP_INSN_SIZE > PAGE_SIZE);
-
 	smp_wmb(); /* pairs with the smp_rmb() in handle_swbp() */
 	set_bit(UPROBE_COPY_INSN, &uprobe->flags);
 
@@ -1166,6 +1162,15 @@ static int __uprobe_register(struct inode *inode, loff_t offset,
 	if (offset > i_size_read(inode))
 		return -EINVAL;
 
+	/*
+	 * This ensures that copy_from_page(), copy_to_page() and
+	 * __update_ref_ctr() can't cross page boundary.
+	 */
+	if (!IS_ALIGNED(offset, UPROBE_SWBP_INSN_SIZE))
+		return -EINVAL;
+	if (!IS_ALIGNED(ref_ctr_offset, sizeof(short)))
+		return -EINVAL;
+
  retry:
 	uprobe = alloc_uprobe(inode, offset, ref_ctr_offset);
 	if (!uprobe)
@@ -2014,6 +2019,9 @@ static int is_trap_at_addr(struct mm_struct *mm, unsigned long vaddr)
 	uprobe_opcode_t opcode;
 	int result;
 
+	if (WARN_ON_ONCE(!IS_ALIGNED(vaddr, UPROBE_SWBP_INSN_SIZE)))
+		return -EINVAL;
+
 	pagefault_disable();
 	result = __get_user(opcode, (uprobe_opcode_t __user *)vaddr);
 	pagefault_enable();
#5 lore raw

Hi,

Quoted 16 lines · Christian Borntraeger
Quoted 3 lines · Oleg Nesterov

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

I think the idea was that the uprobe instruction is 2 bytes and instructions are always aligned to 2 bytes on s390. (we can have 2,4 or 6 bytes).

Quoted 5 lines

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Quoted 4 lines

Not sure if it would have been possible to try to create a uprobe on an odd address. If yes, then the new IS_ALIGNED check certainly makes this better for s390, so the patch looks sane. Adding Vasily and Sven to double check.

I did a quick test, and without this patch it is possible to place a uprobe at an odd address. With the patch it fails with EINVAL, which is more reasonable.

Regards
Sven

#6 lore raw
In reply to Christian Borntraeger
Quoted 10 lines · Christian Borntraeger
Quoted 3 lines · Oleg Nesterov

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

I think the idea was that the uprobe instruction is 2 bytes and instructions are always aligned to 2 bytes on s390. (we can have 2,4 or 6 bytes).

Agree, csky has two length-types of instructions (2,4 bytes).

Quoted 11 lines
Quoted 6 lines

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Not sure if it would have been possible to try to create a uprobe on an odd address. If yes, then the new IS_ALIGNED check certainly makes this better for s390, so the patch looks sane. Adding Vasily and Sven to double check.

Also good to csky.

Signature · 4 lines
Best Regards
 Guo Ren

ML: https://lore.kernel.org/linux-csky/
Reviewed-by on v1 1/1
#7 lore raw
In reply to Oleg Nesterov
Quoted 9 lines · Oleg Nesterov

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Quoted 8 lines

Another problem is __update_ref_ctr() which was wrong from the very beginning, it can read/write outside of kmap'ed page unless "vaddr" is aligned to sizeof(short), __uprobe_register() should check this too.

Thanks Oleg.

Looks good to me.

Quoted 1 line

---

Signature · 2 lines
Thanks and Regards
Srikar Dronamraju
#8 lore raw
Quoted 20 lines · On Wed, 6 May 2020 10:59:55 +0530 Srikar Dronamraju…
Quoted 9 lines · Oleg Nesterov

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Quoted 8 lines

Another problem is __update_ref_ctr() which was wrong from the very beginning, it can read/write outside of kmap'ed page unless "vaddr" is aligned to sizeof(short), __uprobe_register() should check this too.

Quoted 7 lines

Thanks Oleg.

Looks good to me.

Quoted 1 line

---

Thanks Oleg, Srikar and Sven.

As this is in the kernel/events/ directory, I'm guessing it should be taken through the tip tree?

-- Steve

author maintainer
#9 lore raw
Quoted 3 lines · Steven Rostedt

As this is in the kernel/events/ directory, I'm guessing it should be taken through the tip tree?

this would be great, thanks Steven.

Oleg.

author maintainer
#10 lore raw
In reply to Oleg Nesterov

Looks like this patch was forgotten...

Should I resend it?

Quoted 66 lines · Oleg Nesterov

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Another problem is __update_ref_ctr() which was wrong from the very beginning, it can read/write outside of kmap'ed page unless "vaddr" is aligned to sizeof(short), __uprobe_register() should check this too.

---
 kernel/events/uprobes.c | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)
kernel/events/uprobes.c +12 -4
index ece7e13f6e4a..cc2095607c74 100644
@@ -867,10 +867,6 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file,
 	if (ret)
 		goto out;
 
-	/* uprobe_write_opcode() assumes we don't cross page boundary */
-	BUG_ON((uprobe->offset & ~PAGE_MASK) +
-			UPROBE_SWBP_INSN_SIZE > PAGE_SIZE);
-
 	smp_wmb(); /* pairs with the smp_rmb() in handle_swbp() */
 	set_bit(UPROBE_COPY_INSN, &uprobe->flags);
 
@@ -1166,6 +1162,15 @@ static int __uprobe_register(struct inode *inode, loff_t offset,
 	if (offset > i_size_read(inode))
 		return -EINVAL;
 
+	/*
+	 * This ensures that copy_from_page(), copy_to_page() and
+	 * __update_ref_ctr() can't cross page boundary.
+	 */
+	if (!IS_ALIGNED(offset, UPROBE_SWBP_INSN_SIZE))
+		return -EINVAL;
+	if (!IS_ALIGNED(ref_ctr_offset, sizeof(short)))
+		return -EINVAL;
+
  retry:
 	uprobe = alloc_uprobe(inode, offset, ref_ctr_offset);
 	if (!uprobe)
@@ -2014,6 +2019,9 @@ static int is_trap_at_addr(struct mm_struct *mm, unsigned long vaddr)
 	uprobe_opcode_t opcode;
 	int result;
 
+	if (WARN_ON_ONCE(!IS_ALIGNED(vaddr, UPROBE_SWBP_INSN_SIZE)))
+		return -EINVAL;
+
 	pagefault_disable();
 	result = __get_user(opcode, (uprobe_opcode_t __user *)vaddr);
 	pagefault_enable();
Signature · 1 line
2.25.1.362.g51ebf55
#11 lore raw
Quoted 4 lines · Oleg Nesterov

Looks like this patch was forgotten...

Should I resend it?

I guess I'll just take it directly, since it was triggered by me complaining anyway.

I had hoped it would go through the usual channels.

Linus

#12 lore raw
Quoted 10 lines · On Tue, 9 Jun 2020 09:48:45 -0700 Linus Torvalds…
Quoted 4 lines · Oleg Nesterov

Looks like this patch was forgotten...

Should I resend it?

I guess I'll just take it directly, since it was triggered by me complaining anyway.

I had hoped it would go through the usual channels.

Perhaps there was confusion about which tree it was suppose to go through :-/

-- Steve

Reviewed-by on v1 1/1
#13 lore raw
In reply to Oleg Nesterov

Hi Oleg,

Quoted 64 lines · Oleg Nesterov

uprobe_write_opcode() must not cross page boundary; prepare_uprobe() relies on arch_uprobe_analyze_insn() which should validate "vaddr" but some architectures (csky, s390, and sparc) don't do this.

We can remove the BUG_ON() check in prepare_uprobe() and validate the offset early in __uprobe_register(). The new IS_ALIGNED() check matches the alignment check in arch_prepare_kprobe() on supported architectures, so I think that all insns must be aligned to UPROBE_SWBP_INSN_SIZE.

Another problem is __update_ref_ctr() which was wrong from the very beginning, it can read/write outside of kmap'ed page unless "vaddr" is aligned to sizeof(short), __uprobe_register() should check this too.

---
 kernel/events/uprobes.c | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)
kernel/events/uprobes.c +12 -5
index ece7e13f6e4a..cc2095607c74 100644
@@ -867,10 +867,6 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file,
        if (ret)
                goto out;
 
-       /* uprobe_write_opcode() assumes we don't cross page boundary */
-       BUG_ON((uprobe->offset & ~PAGE_MASK) +
-                       UPROBE_SWBP_INSN_SIZE > PAGE_SIZE);
-
        smp_wmb(); /* pairs with the smp_rmb() in handle_swbp() */
        set_bit(UPROBE_COPY_INSN, &uprobe->flags);
 
@@ -1166,6 +1162,15 @@ static int __uprobe_register(struct inode *inode, loff_t offset,
        if (offset > i_size_read(inode))
                return -EINVAL;
 
+       /*
+        * This ensures that copy_from_page(), copy_to_page() and
+        * __update_ref_ctr() can't cross page boundary.
+        */
+       if (!IS_ALIGNED(offset, UPROBE_SWBP_INSN_SIZE))
+               return -EINVAL;
+       if (!IS_ALIGNED(ref_ctr_offset, sizeof(short)))
+               return -EINVAL;
+
  retry:
        uprobe = alloc_uprobe(inode, offset, ref_ctr_offset);
        if (!uprobe)
@@ -2014,6 +2019,9 @@ static int is_trap_at_addr(struct mm_struct *mm, unsigned long vaddr)
        uprobe_opcode_t opcode;
        int result;
 
+       if (WARN_ON_ONCE(!IS_ALIGNED(vaddr, UPROBE_SWBP_INSN_SIZE)))
+               return -EINVAL;
+
        pagefault_disable();
        result = __get_user(opcode, (uprobe_opcode_t __user *)vaddr);
        pagefault_enable();
--
Quoted 3 lines

2.25.1.362.g51ebf55

Looks good to me, thx.

Signature · 4 lines
Best Regards
 Guo Ren

ML: https://lore.kernel.org/linux-csky/