[stable,6.6,v2] lib/buildid: Handle memfd_secret() files in build_id_parse()

Commit Message

Chen Linxuan March 17, 2025, 1:16 a.m. UTC

[ Upstream commit 5ac9b4e935dfc6af41eee2ddc21deb5c36507a9f ]

>From memfd_secret(2) manpage:

  The memory areas backing the file created with memfd_secret(2) are
  visible only to the processes that have access to the file descriptor.
  The memory region is removed from the kernel page tables and only the
  page tables of the processes holding the file descriptor map the
  corresponding physical memory. (Thus, the pages in the region can't be
  accessed by the kernel itself, so that, for example, pointers to the
  region can't be passed to system calls.)

We need to handle this special case gracefully in build ID fetching
code. Return -EFAULT whenever secretmem file is passed to build_id_parse()
family of APIs. Original report and repro can be found in [0].

  [0] https://lore.kernel.org/bpf/ZwyG8Uro%2FSyTXAni@ly-workstation/

Fixes: de3ec364c3c3 ("lib/buildid: add single folio-based file reader abstraction")
Reported-by: Yi Lai <[e-mail]>
Suggested-by: Shakeel Butt <[e-mail]>
Signed-off-by: Andrii Nakryiko <[e-mail]>
Signed-off-by: Daniel Borkmann <[e-mail]>
Acked-by: Shakeel Butt <[e-mail]>
Link: https://lore.kernel.org/bpf/[e-mail]
Link: https://lore.kernel.org/bpf/[e-mail]
[ Chen Linxuan: backport same logic without folio-based changes ]
Cc: [e-mail]
Fixes: 88a16a130933 ("perf: Add build id data in mmap2 event")
Signed-off-by: Chen Linxuan <[e-mail]>
---
v1 -> v2: use vma_is_secretmem() instead of directly checking
          vma->vm_file->f_op == &secretmem_fops
---
 lib/buildid.c | 5 +++++
 1 file changed, 5 insertions(+)

Comments

Greg KH March 17, 2025, 4:18 a.m. UTC | #1

On Mon, Mar 17, 2025 at 09:16:04AM +0800, Chen Linxuan wrote:
> [ Upstream commit 5ac9b4e935dfc6af41eee2ddc21deb5c36507a9f ]
> 
> >>From memfd_secret(2) manpage:
> 
>   The memory areas backing the file created with memfd_secret(2) are
>   visible only to the processes that have access to the file descriptor.
>   The memory region is removed from the kernel page tables and only the
>   page tables of the processes holding the file descriptor map the
>   corresponding physical memory. (Thus, the pages in the region can't be
>   accessed by the kernel itself, so that, for example, pointers to the
>   region can't be passed to system calls.)
> 
> We need to handle this special case gracefully in build ID fetching
> code. Return -EFAULT whenever secretmem file is passed to build_id_parse()
> family of APIs. Original report and repro can be found in [0].
> 
>   [0] https://lore.kernel.org/bpf/ZwyG8Uro%2FSyTXAni@ly-workstation/
> 
> Fixes: de3ec364c3c3 ("lib/buildid: add single folio-based file reader abstraction")
> Reported-by: Yi Lai <[e-mail]>
> Suggested-by: Shakeel Butt <[e-mail]>
> Signed-off-by: Andrii Nakryiko <[e-mail]>
> Signed-off-by: Daniel Borkmann <[e-mail]>
> Acked-by: Shakeel Butt <[e-mail]>
> Link: https://lore.kernel.org/bpf/[e-mail]
> Link: https://lore.kernel.org/bpf/[e-mail]
> [ Chen Linxuan: backport same logic without folio-based changes ]
> Cc: [e-mail]
> Fixes: 88a16a130933 ("perf: Add build id data in mmap2 event")
> Signed-off-by: Chen Linxuan <[e-mail]>
> ---
> v1 -> v2: use vma_is_secretmem() instead of directly checking
>           vma->vm_file->f_op == &secretmem_fops
> ---
>  lib/buildid.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/lib/buildid.c b/lib/buildid.c
> index 9fc46366597e..34315d09b544 100644
> --- a/lib/buildid.c
> +++ b/lib/buildid.c
> @@ -5,6 +5,7 @@
>  #include <linux/elf.h>
>  #include <linux/kernel.h>
>  #include <linux/pagemap.h>
> +#include <linux/secretmem.h>
>  
>  #define BUILD_ID 3
>  
> @@ -157,6 +158,10 @@ int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
>  	if (!vma->vm_file)
>  		return -EINVAL;
>  
> +	/* reject secretmem */

Why is this comment different from what is in the original commit?  Same
for your other backports.  Please try to keep it as identical to the
original whenever possible as we have to maintain this for a very long
time.

thanks,

greg k-h

Chen Linxuan March 17, 2025, 5:04 a.m. UTC | #2

Greg KH <[e-mail]> 于2025年3月17日周一 12:20写道：
>
> On Mon, Mar 17, 2025 at 09:16:04AM +0800, Chen Linxuan wrote:
> > [ Upstream commit 5ac9b4e935dfc6af41eee2ddc21deb5c36507a9f ]
> >
> > >>From memfd_secret(2) manpage:
> >
> >   The memory areas backing the file created with memfd_secret(2) are
> >   visible only to the processes that have access to the file descriptor.
> >   The memory region is removed from the kernel page tables and only the
> >   page tables of the processes holding the file descriptor map the
> >   corresponding physical memory. (Thus, the pages in the region can't be
> >   accessed by the kernel itself, so that, for example, pointers to the
> >   region can't be passed to system calls.)
> >
> > We need to handle this special case gracefully in build ID fetching
> > code. Return -EFAULT whenever secretmem file is passed to build_id_parse()
> > family of APIs. Original report and repro can be found in [0].
> >
> >   [0] https://lore.kernel.org/bpf/ZwyG8Uro%2FSyTXAni@ly-workstation/
> >
> > Fixes: de3ec364c3c3 ("lib/buildid: add single folio-based file reader abstraction")
> > Reported-by: Yi Lai <[e-mail]>
> > Suggested-by: Shakeel Butt <[e-mail]>
> > Signed-off-by: Andrii Nakryiko <[e-mail]>
> > Signed-off-by: Daniel Borkmann <[e-mail]>
> > Acked-by: Shakeel Butt <[e-mail]>
> > Link: https://lore.kernel.org/bpf/[e-mail]
> > Link: https://lore.kernel.org/bpf/[e-mail]
> > [ Chen Linxuan: backport same logic without folio-based changes ]
> > Cc: [e-mail]
> > Fixes: 88a16a130933 ("perf: Add build id data in mmap2 event")
> > Signed-off-by: Chen Linxuan <[e-mail]>
> > ---
> > v1 -> v2: use vma_is_secretmem() instead of directly checking
> >           vma->vm_file->f_op == &secretmem_fops
> > ---
> >  lib/buildid.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> >
> > diff --git a/lib/buildid.c b/lib/buildid.c
> > index 9fc46366597e..34315d09b544 100644
> > --- a/lib/buildid.c
> > +++ b/lib/buildid.c
> > @@ -5,6 +5,7 @@
> >  #include <linux/elf.h>
> >  #include <linux/kernel.h>
> >  #include <linux/pagemap.h>
> > +#include <linux/secretmem.h>
> >
> >  #define BUILD_ID 3
> >
> > @@ -157,6 +158,10 @@ int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
> >       if (!vma->vm_file)
> >               return -EINVAL;
> >
> > +     /* reject secretmem */
>
> Why is this comment different from what is in the original commit?  Same
> for your other backports.  Please try to keep it as identical to the
> original whenever possible as we have to maintain this for a very long
> time.
>
> thanks,
>
> greg k-h
>
>

Original comment is in a function named freader_get_folio(),
but folio related changes has not been backported yet.

thanks,

Chen Linxuan

Greg Kroah-Hartman March 17, 2025, 5:14 a.m. UTC | #3

On Mon, Mar 17, 2025 at 01:04:41PM +0800, Chen Linxuan wrote:
> Greg KH <[e-mail]> 于2025年3月17日周一 12:20写道：
> >
> > On Mon, Mar 17, 2025 at 09:16:04AM +0800, Chen Linxuan wrote:
> > > [ Upstream commit 5ac9b4e935dfc6af41eee2ddc21deb5c36507a9f ]
> > >
> > > >>From memfd_secret(2) manpage:
> > >
> > >   The memory areas backing the file created with memfd_secret(2) are
> > >   visible only to the processes that have access to the file descriptor.
> > >   The memory region is removed from the kernel page tables and only the
> > >   page tables of the processes holding the file descriptor map the
> > >   corresponding physical memory. (Thus, the pages in the region can't be
> > >   accessed by the kernel itself, so that, for example, pointers to the
> > >   region can't be passed to system calls.)
> > >
> > > We need to handle this special case gracefully in build ID fetching
> > > code. Return -EFAULT whenever secretmem file is passed to build_id_parse()
> > > family of APIs. Original report and repro can be found in [0].
> > >
> > >   [0] https://lore.kernel.org/bpf/ZwyG8Uro%2FSyTXAni@ly-workstation/
> > >
> > > Fixes: de3ec364c3c3 ("lib/buildid: add single folio-based file reader abstraction")
> > > Reported-by: Yi Lai <[e-mail]>
> > > Suggested-by: Shakeel Butt <[e-mail]>
> > > Signed-off-by: Andrii Nakryiko <[e-mail]>
> > > Signed-off-by: Daniel Borkmann <[e-mail]>
> > > Acked-by: Shakeel Butt <[e-mail]>
> > > Link: https://lore.kernel.org/bpf/[e-mail]
> > > Link: https://lore.kernel.org/bpf/[e-mail]
> > > [ Chen Linxuan: backport same logic without folio-based changes ]
> > > Cc: [e-mail]
> > > Fixes: 88a16a130933 ("perf: Add build id data in mmap2 event")
> > > Signed-off-by: Chen Linxuan <[e-mail]>
> > > ---
> > > v1 -> v2: use vma_is_secretmem() instead of directly checking
> > >           vma->vm_file->f_op == &secretmem_fops
> > > ---
> > >  lib/buildid.c | 5 +++++
> > >  1 file changed, 5 insertions(+)
> > >
> > > diff --git a/lib/buildid.c b/lib/buildid.c
> > > index 9fc46366597e..34315d09b544 100644
> > > --- a/lib/buildid.c
> > > +++ b/lib/buildid.c
> > > @@ -5,6 +5,7 @@
> > >  #include <linux/elf.h>
> > >  #include <linux/kernel.h>
> > >  #include <linux/pagemap.h>
> > > +#include <linux/secretmem.h>
> > >
> > >  #define BUILD_ID 3
> > >
> > > @@ -157,6 +158,10 @@ int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
> > >       if (!vma->vm_file)
> > >               return -EINVAL;
> > >
> > > +     /* reject secretmem */
> >
> > Why is this comment different from what is in the original commit?  Same
> > for your other backports.  Please try to keep it as identical to the
> > original whenever possible as we have to maintain this for a very long
> > time.
> >
> > thanks,
> >
> > greg k-h
> >
> >
> 
> Original comment is in a function named freader_get_folio(),
> but folio related changes has not been backported yet.

That's fine, but the logic is the same so keep the original code as
close as possible.  Otherwise it looks like this is a totally different
change and we would have to reject it for obvious reasons.

thanks,

greg k-h

Chen Linxuan March 17, 2025, 5:38 a.m. UTC | #4

Greg KH <[e-mail]> 于2025年3月17日周一 13:15写道：
>
> On Mon, Mar 17, 2025 at 01:04:41PM +0800, Chen Linxuan wrote:
> > Greg KH <[e-mail]> 于2025年3月17日周一 12:20写道：
> > >
> > > On Mon, Mar 17, 2025 at 09:16:04AM +0800, Chen Linxuan wrote:
> > > > [ Upstream commit 5ac9b4e935dfc6af41eee2ddc21deb5c36507a9f ]
> > > >
> > > > >>From memfd_secret(2) manpage:
> > > >
> > > >   The memory areas backing the file created with memfd_secret(2) are
> > > >   visible only to the processes that have access to the file descriptor.
> > > >   The memory region is removed from the kernel page tables and only the
> > > >   page tables of the processes holding the file descriptor map the
> > > >   corresponding physical memory. (Thus, the pages in the region can't be
> > > >   accessed by the kernel itself, so that, for example, pointers to the
> > > >   region can't be passed to system calls.)
> > > >
> > > > We need to handle this special case gracefully in build ID fetching
> > > > code. Return -EFAULT whenever secretmem file is passed to build_id_parse()
> > > > family of APIs. Original report and repro can be found in [0].
> > > >
> > > >   [0] https://lore.kernel.org/bpf/ZwyG8Uro%2FSyTXAni@ly-workstation/
> > > >
> > > > Fixes: de3ec364c3c3 ("lib/buildid: add single folio-based file reader abstraction")
> > > > Reported-by: Yi Lai <[e-mail]>
> > > > Suggested-by: Shakeel Butt <[e-mail]>
> > > > Signed-off-by: Andrii Nakryiko <[e-mail]>
> > > > Signed-off-by: Daniel Borkmann <[e-mail]>
> > > > Acked-by: Shakeel Butt <[e-mail]>
> > > > Link: https://lore.kernel.org/bpf/[e-mail]
> > > > Link: https://lore.kernel.org/bpf/[e-mail]
> > > > [ Chen Linxuan: backport same logic without folio-based changes ]
> > > > Cc: [e-mail]
> > > > Fixes: 88a16a130933 ("perf: Add build id data in mmap2 event")
> > > > Signed-off-by: Chen Linxuan <[e-mail]>
> > > > ---
> > > > v1 -> v2: use vma_is_secretmem() instead of directly checking
> > > >           vma->vm_file->f_op == &secretmem_fops
> > > > ---
> > > >  lib/buildid.c | 5 +++++
> > > >  1 file changed, 5 insertions(+)
> > > >
> > > > diff --git a/lib/buildid.c b/lib/buildid.c
> > > > index 9fc46366597e..34315d09b544 100644
> > > > --- a/lib/buildid.c
> > > > +++ b/lib/buildid.c
> > > > @@ -5,6 +5,7 @@
> > > >  #include <linux/elf.h>
> > > >  #include <linux/kernel.h>
> > > >  #include <linux/pagemap.h>
> > > > +#include <linux/secretmem.h>
> > > >
> > > >  #define BUILD_ID 3
> > > >
> > > > @@ -157,6 +158,10 @@ int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
> > > >       if (!vma->vm_file)
> > > >               return -EINVAL;
> > > >
> > > > +     /* reject secretmem */
> > >
> > > Why is this comment different from what is in the original commit?  Same
> > > for your other backports.  Please try to keep it as identical to the
> > > original whenever possible as we have to maintain this for a very long
> > > time.
> > >
> > > thanks,
> > >
> > > greg k-h
> > >
> > >
> >
> > Original comment is in a function named freader_get_folio(),
> > but folio related changes has not been backported yet.
>
> That's fine, but the logic is the same so keep the original code as
> close as possible.  Otherwise it looks like this is a totally different
> change and we would have to reject it for obvious reasons.
>
> thanks,
>
> greg k-h
>
>

V3 has been sent.

Thanks,
Chen Linxuan

Patch

diff --git a/lib/buildid.c b/lib/buildid.c
index 9fc46366597e..34315d09b544 100644
--- a/lib/buildid.c
+++ b/lib/buildid.c
@@ -5,6 +5,7 @@ 
 #include <linux/elf.h>
 #include <linux/kernel.h>
 #include <linux/pagemap.h>
+#include <linux/secretmem.h>
 
 #define BUILD_ID 3
 
@@ -157,6 +158,10 @@  int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id,
 	if (!vma->vm_file)
 		return -EINVAL;
 
+	/* reject secretmem */
+	if (vma_is_secretmem(vma))
+		return -EFAULT;
+
 	page = find_get_page(vma->vm_file->f_mapping, 0);
 	if (!page)
 		return -EFAULT;	/* page not mapped */