[PATCH mm] maccess: fix strncpy_from_user_nofault empty string handling

[PATCH mm] maccess: fix strncpy_from_user_nofault empty string handling

[Mykyta Yatsenko]

From: Mykyta Yatsenko <yatsenko@meta.com>

strncpy_from_user_nofault should return the length of the copied string
including the trailing NUL, but if the argument unsafe_addr points to
an empty string ({'\0'}), the return value is 0.

This happens as strncpy_from_user copies terminal symbol into dst
and returns 0 (as expected), but strncpy_from_user_nofault does not
modify ret as it is not equal to count and not greater than 0, so 0 is
returned, which contradicts the contract.

Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>
---
 kernel/trace/trace_events_filter.c | 10 ++++++++--
 mm/maccess.c                       |  2 +-
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
index 0993dfc1c5c1..86b7e5a4e235 100644
--- a/kernel/trace/trace_events_filter.c
+++ b/kernel/trace/trace_events_filter.c
@@ -800,6 +800,7 @@ static __always_inline char *test_string(char *str)
 {
 	struct ustring_buffer *ubuf;
 	char *kstr;
+	int cnt;
 
 	if (!ustring_per_cpu)
 		return NULL;
@@ -808,7 +809,9 @@ static __always_inline char *test_string(char *str)
 	kstr = ubuf->buffer;
 
 	/* For safety, do not trust the string pointer */
-	if (!strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE))
+	cnt = strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE);
+	/* Return null if empty string or error */
+	if (cnt <= 1)
 		return NULL;
 	return kstr;
 }
@@ -818,6 +821,7 @@ static __always_inline char *test_ustring(char *str)
 	struct ustring_buffer *ubuf;
 	char __user *ustr;
 	char *kstr;
+	int cnt;
 
 	if (!ustring_per_cpu)
 		return NULL;
@@ -827,7 +831,9 @@ static __always_inline char *test_ustring(char *str)
 
 	/* user space address? */
 	ustr = (char __user *)str;
-	if (!strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE))
+	cnt = strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE);
+	/* Return null if empty string or error */
+	if (cnt <= 1)
 		return NULL;
 
 	return kstr;
diff --git a/mm/maccess.c b/mm/maccess.c
index 8f0906180a94..831b4dd7296c 100644
--- a/mm/maccess.c
+++ b/mm/maccess.c
@@ -196,7 +196,7 @@ long strncpy_from_user_nofault(char *dst, const void __user *unsafe_addr,
 	if (ret >= count) {
 		ret = count;
 		dst[ret - 1] = '\0';
-	} else if (ret > 0) {
+	} else if (ret >= 0) {
 		ret++;
 	}
 
-- 
2.49.0

Re: [PATCH mm] maccess: fix strncpy_from_user_nofault empty string handling

[Andrew Morton]

On Thu, 17 Apr 2025 16:28:08 +0100 Mykyta Yatsenko <mykyta.yatsenko5@gmail.com> wrote:

> strncpy_from_user_nofault should return the length of the copied string
> including the trailing NUL, but if the argument unsafe_addr points to
> an empty string ({'\0'}), the return value is 0.
> 
> This happens as strncpy_from_user copies terminal symbol into dst
> and returns 0 (as expected), but strncpy_from_user_nofault does not
> modify ret as it is not equal to count and not greater than 0, so 0 is
> returned, which contradicts the contract.

Looks right, I think.

But why do strncpy_from_user() and strncpy_from_user_nofault() have
different interfaces?

/**
 * strncpy_from_user: - Copy a NUL terminated string from userspace.
 * ...
 * On success, returns the length of the string (not including the trailing
 * NUL).

/**
 * strncpy_from_user_nofault: - Copy a NUL terminated string from unsafe user
 *				address.
 * ...
 * On success, returns the length of the string INCLUDING the trailing NUL.

This is surprising and I'm wondering what led us to do this?

Re: [PATCH mm] maccess: fix strncpy_from_user_nofault empty string handling

[Andrii Nakryiko]

On Thu, Apr 17, 2025 at 8:28 AM Mykyta Yatsenko
<mykyta.yatsenko5@gmail.com> wrote:
>
> From: Mykyta Yatsenko <yatsenko@meta.com>
>
> strncpy_from_user_nofault should return the length of the copied string
> including the trailing NUL, but if the argument unsafe_addr points to
> an empty string ({'\0'}), the return value is 0.
>
> This happens as strncpy_from_user copies terminal symbol into dst
> and returns 0 (as expected), but strncpy_from_user_nofault does not
> modify ret as it is not equal to count and not greater than 0, so 0 is
> returned, which contradicts the contract.
>
> Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>
> ---
>  kernel/trace/trace_events_filter.c | 10 ++++++++--
>  mm/maccess.c                       |  2 +-
>  2 files changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
> index 0993dfc1c5c1..86b7e5a4e235 100644
> --- a/kernel/trace/trace_events_filter.c
> +++ b/kernel/trace/trace_events_filter.c
> @@ -800,6 +800,7 @@ static __always_inline char *test_string(char *str)
>  {
>         struct ustring_buffer *ubuf;
>         char *kstr;
> +       int cnt;
>
>         if (!ustring_per_cpu)
>                 return NULL;
> @@ -808,7 +809,9 @@ static __always_inline char *test_string(char *str)
>         kstr = ubuf->buffer;
>
>         /* For safety, do not trust the string pointer */
> -       if (!strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE))
> +       cnt = strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE);
> +       /* Return null if empty string or error */
> +       if (cnt <= 1)
>                 return NULL;

I wouldn't touch this part and leave it up to Steven to fix (if he
agrees it needs fixing). Current logic seems wrong already, as it
won't correctly handle -EFAULT. And, on the other hand, there is
nothing wrong or special about empty string, so I don't think it needs
special handling. Let's drop these changes in trace_events_filter.c?

>         return kstr;
>  }
> @@ -818,6 +821,7 @@ static __always_inline char *test_ustring(char *str)
>         struct ustring_buffer *ubuf;
>         char __user *ustr;
>         char *kstr;
> +       int cnt;
>
>         if (!ustring_per_cpu)
>                 return NULL;
> @@ -827,7 +831,9 @@ static __always_inline char *test_ustring(char *str)
>
>         /* user space address? */
>         ustr = (char __user *)str;
> -       if (!strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE))
> +       cnt = strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE);
> +       /* Return null if empty string or error */
> +       if (cnt <= 1)
>                 return NULL;

ditto

>
>         return kstr;
> diff --git a/mm/maccess.c b/mm/maccess.c
> index 8f0906180a94..831b4dd7296c 100644
> --- a/mm/maccess.c
> +++ b/mm/maccess.c
> @@ -196,7 +196,7 @@ long strncpy_from_user_nofault(char *dst, const void __user *unsafe_addr,
>         if (ret >= count) {
>                 ret = count;
>                 dst[ret - 1] = '\0';
> -       } else if (ret > 0) {
> +       } else if (ret >= 0) {
>                 ret++;
>         }
>

This part looks good and does indeed fix the issue. Good catch!

Reviewed-by: Andrii Nakryiko <andrii@kernel.org>

> --
> 2.49.0
>

Re: [PATCH mm] maccess: fix strncpy_from_user_nofault empty string handling

[Andrii Nakryiko]

On Thu, Apr 17, 2025 at 1:44 PM Andrii Nakryiko
<andrii.nakryiko@gmail.com> wrote:
>
> On Thu, Apr 17, 2025 at 8:28 AM Mykyta Yatsenko
> <mykyta.yatsenko5@gmail.com> wrote:
> >
> > From: Mykyta Yatsenko <yatsenko@meta.com>
> >
> > strncpy_from_user_nofault should return the length of the copied string
> > including the trailing NUL, but if the argument unsafe_addr points to
> > an empty string ({'\0'}), the return value is 0.
> >
> > This happens as strncpy_from_user copies terminal symbol into dst
> > and returns 0 (as expected), but strncpy_from_user_nofault does not
> > modify ret as it is not equal to count and not greater than 0, so 0 is
> > returned, which contradicts the contract.
> >
> > Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>
> > ---
> >  kernel/trace/trace_events_filter.c | 10 ++++++++--
> >  mm/maccess.c                       |  2 +-
> >  2 files changed, 9 insertions(+), 3 deletions(-)
> >
> > diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
> > index 0993dfc1c5c1..86b7e5a4e235 100644
> > --- a/kernel/trace/trace_events_filter.c
> > +++ b/kernel/trace/trace_events_filter.c
> > @@ -800,6 +800,7 @@ static __always_inline char *test_string(char *str)
> >  {
> >         struct ustring_buffer *ubuf;
> >         char *kstr;
> > +       int cnt;
> >
> >         if (!ustring_per_cpu)
> >                 return NULL;
> > @@ -808,7 +809,9 @@ static __always_inline char *test_string(char *str)
> >         kstr = ubuf->buffer;
> >
> >         /* For safety, do not trust the string pointer */
> > -       if (!strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE))
> > +       cnt = strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE);
> > +       /* Return null if empty string or error */
> > +       if (cnt <= 1)
> >                 return NULL;
>
> I wouldn't touch this part and leave it up to Steven to fix (if he
> agrees it needs fixing). Current logic seems wrong already, as it
> won't correctly handle -EFAULT. And, on the other hand, there is
> nothing wrong or special about empty string, so I don't think it needs
> special handling. Let's drop these changes in trace_events_filter.c?

Actually, you are not even touching strncpy_from_kernel_nofault(), so
yeah, definitely let's not do this (at least not in this patch).

>
> >         return kstr;
> >  }
> > @@ -818,6 +821,7 @@ static __always_inline char *test_ustring(char *str)
> >         struct ustring_buffer *ubuf;
> >         char __user *ustr;
> >         char *kstr;
> > +       int cnt;
> >
> >         if (!ustring_per_cpu)
> >                 return NULL;
> > @@ -827,7 +831,9 @@ static __always_inline char *test_ustring(char *str)
> >
> >         /* user space address? */
> >         ustr = (char __user *)str;
> > -       if (!strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE))
> > +       cnt = strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE);
> > +       /* Return null if empty string or error */
> > +       if (cnt <= 1)
> >                 return NULL;
>
> ditto
>
> >
> >         return kstr;
> > diff --git a/mm/maccess.c b/mm/maccess.c
> > index 8f0906180a94..831b4dd7296c 100644
> > --- a/mm/maccess.c
> > +++ b/mm/maccess.c
> > @@ -196,7 +196,7 @@ long strncpy_from_user_nofault(char *dst, const void __user *unsafe_addr,
> >         if (ret >= count) {
> >                 ret = count;
> >                 dst[ret - 1] = '\0';
> > -       } else if (ret > 0) {
> > +       } else if (ret >= 0) {
> >                 ret++;
> >         }
> >
>
> This part looks good and does indeed fix the issue. Good catch!
>
> Reviewed-by: Andrii Nakryiko <andrii@kernel.org>
>
> > --
> > 2.49.0
> >

Re: [PATCH mm] maccess: fix strncpy_from_user_nofault empty string handling

[Andrii Nakryiko]

On Thu, Apr 17, 2025 at 1:40 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Thu, 17 Apr 2025 16:28:08 +0100 Mykyta Yatsenko <mykyta.yatsenko5@gmail.com> wrote:
>
> > strncpy_from_user_nofault should return the length of the copied string
> > including the trailing NUL, but if the argument unsafe_addr points to
> > an empty string ({'\0'}), the return value is 0.
> >
> > This happens as strncpy_from_user copies terminal symbol into dst
> > and returns 0 (as expected), but strncpy_from_user_nofault does not
> > modify ret as it is not equal to count and not greater than 0, so 0 is
> > returned, which contradicts the contract.
>
> Looks right, I think.
>
> But why do strncpy_from_user() and strncpy_from_user_nofault() have
> different interfaces?
>
> /**
>  * strncpy_from_user: - Copy a NUL terminated string from userspace.
>  * ...
>  * On success, returns the length of the string (not including the trailing
>  * NUL).
>
> /**
>  * strncpy_from_user_nofault: - Copy a NUL terminated string from unsafe user
>  *                              address.
>  * ...
>  * On success, returns the length of the string INCLUDING the trailing NUL.
>
> This is surprising and I'm wondering what led us to do this?

Agreed, this is very surprising and error-prone. strncpy_from_user()
semantics is a bit better, IMO, in that it allows to "detect" empty
string even if buffer size is 1 byte. And there isn't a lot of places
where we use strncpy_from_user_nofault (only 6, it seems). Maybe we
should just change the semantics of strncpy_from_user_nofault?

Re: [PATCH mm] maccess: fix strncpy_from_user_nofault empty string handling

[Steven Rostedt]

On Thu, 17 Apr 2025 13:44:48 -0700
Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:

> > @@ -808,7 +809,9 @@ static __always_inline char *test_string(char *str)
> >         kstr = ubuf->buffer;
> >
> >         /* For safety, do not trust the string pointer */
> > -       if (!strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE))
> > +       cnt = strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE);
> > +       /* Return null if empty string or error */
> > +       if (cnt <= 1)
> >                 return NULL;  
> 
> I wouldn't touch this part and leave it up to Steven to fix (if he
> agrees it needs fixing). Current logic seems wrong already, as it
> won't correctly handle -EFAULT. And, on the other hand, there is
> nothing wrong or special about empty string, so I don't think it needs
> special handling. Let's drop these changes in trace_events_filter.c?

Bah, it is wrong. I don't usually use filtering on strings much, but come
to think of it, the last time I tried, it didn't work, but I found another
way to get what I was looking for, and didn't look deeper into it.

I only care if it faulted or not. I don't care about it just copying zero
bytes. It should have been:

	if (strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE) < 0)

> 
> >         return kstr;
> >  }
> > @@ -818,6 +821,7 @@ static __always_inline char *test_ustring(char *str)
> >         struct ustring_buffer *ubuf;
> >         char __user *ustr;
> >         char *kstr;
> > +       int cnt;
> >
> >         if (!ustring_per_cpu)
> >                 return NULL;
> > @@ -827,7 +831,9 @@ static __always_inline char *test_ustring(char *str)
> >
> >         /* user space address? */
> >         ustr = (char __user *)str;
> > -       if (!strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE))

This is broken too.

As this isn't relying on the other change in this patch, I'll just fix it
myself. I'm getting a pull request ready anyway.

Thanks!

-- Steve


> > +       cnt = strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE);
> > +       /* Return null if empty string or error */
> > +       if (cnt <= 1)
> >                 return NULL;  
> 
> ditto
> 
> >