x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

[PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Harshit Mogalapalli 2 months, 4 weeks ago

When the second-stage kernel is booted via kexec with a limiting command
line such as "mem=<size>", the physical range that contains the carried
over IMA measurement list may fall outside the truncated RAM leading to
a kernel panic.

    BUG: unable to handle page fault for address: ffff97793ff47000
    RIP: ima_restore_measurement_list+0xdc/0x45a
    #PF: error_code(0x0000) – not-present page

Other architectures already validate the range with page_is_ram(), as
done in commit: cbf9c4b9617b ("of: check previous kernel's
ima-kexec-buffer against memory bounds") do a similar check on x86.

Cc: stable@vger.kernel.org
Fixes: b69a2afd5afc ("x86/kexec: Carry forward IMA measurement log on kexec")
Reported-by: Paul Webb <paul.x.webb@oracle.com>
Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
---
Have tested the kexec for x86 kernel with IMA_KEXEC enabled and the
above patch works good. Paul initially reported this on 6.12 kernel but
I was able to reproduce this on 6.18, so I tried replicating how this
was fixed in drivers/of/kexec.c
---
 arch/x86/kernel/setup.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index 1b2edd07a3e1..fcef197d180e 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -439,9 +439,23 @@ int __init ima_free_kexec_buffer(void)
 
 int __init ima_get_kexec_buffer(void **addr, size_t *size)
 {
+	unsigned long start_pfn, end_pfn;
+
 	if (!ima_kexec_buffer_size)
 		return -ENOENT;
 
+	/*
+	 * Calculate the PFNs for the buffer and ensure
+	 * they are with in addressable memory.
+	 */
+	start_pfn = PFN_DOWN(ima_kexec_buffer_phys);
+	end_pfn = PFN_DOWN(ima_kexec_buffer_phys + ima_kexec_buffer_size - 1);
+	if (!pfn_range_is_mapped(start_pfn, end_pfn)) {
+		pr_warn("IMA buffer at 0x%llx, size = 0x%zx beyond memory\n",
+			ima_kexec_buffer_phys, ima_kexec_buffer_size);
+		return -EINVAL;
+	}
+
 	*addr = __va(ima_kexec_buffer_phys);
 	*size = ima_kexec_buffer_size;
 
-- 
2.50.1

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Andrew Morton 2 months, 1 week ago

On Wed, 12 Nov 2025 11:30:02 -0800 Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com> wrote:

> When the second-stage kernel is booted via kexec with a limiting command
> line such as "mem=<size>", the physical range that contains the carried
> over IMA measurement list may fall outside the truncated RAM leading to
> a kernel panic.
> 
>     BUG: unable to handle page fault for address: ffff97793ff47000
>     RIP: ima_restore_measurement_list+0xdc/0x45a
>     #PF: error_code(0x0000) – not-present page
> 
> Other architectures already validate the range with page_is_ram(), as
> done in commit: cbf9c4b9617b ("of: check previous kernel's
> ima-kexec-buffer against memory bounds") do a similar check on x86.

Thanks.

> Cc: stable@vger.kernel.org
> Fixes: b69a2afd5afc ("x86/kexec: Carry forward IMA measurement log on kexec")

That was via the x86 tree so I assume the x86 team (Boris?) will be
processing this patch.

I'll put it into mm.git's mm-hotfixes branch for now, to get a bit of
testing and to generally track its progress.

> --- a/arch/x86/kernel/setup.c
> +++ b/arch/x86/kernel/setup.c
> @@ -439,9 +439,23 @@ int __init ima_free_kexec_buffer(void)
>  
>  int __init ima_get_kexec_buffer(void **addr, size_t *size)
>  {
> +	unsigned long start_pfn, end_pfn;
> +
>  	if (!ima_kexec_buffer_size)
>  		return -ENOENT;
>  
> +	/*
> +	 * Calculate the PFNs for the buffer and ensure
> +	 * they are with in addressable memory.

"within" ;)

> +	 */
> +	start_pfn = PFN_DOWN(ima_kexec_buffer_phys);
> +	end_pfn = PFN_DOWN(ima_kexec_buffer_phys + ima_kexec_buffer_size - 1);
> +	if (!pfn_range_is_mapped(start_pfn, end_pfn)) {
> +		pr_warn("IMA buffer at 0x%llx, size = 0x%zx beyond memory\n",
> +			ima_kexec_buffer_phys, ima_kexec_buffer_size);
> +		return -EINVAL;
> +	}
> +
>  	*addr = __va(ima_kexec_buffer_phys);
>  	*size = ima_kexec_buffer_size;

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Borislav Petkov 2 months, 1 week ago

On Mon, Dec 01, 2025 at 09:20:20AM -0800, Andrew Morton wrote:
> On Wed, 12 Nov 2025 11:30:02 -0800 Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com> wrote:
> 
> > When the second-stage kernel is booted via kexec with a limiting command
> > line such as "mem=<size>", the physical range that contains the carried
> > over IMA measurement list may fall outside the truncated RAM leading to
> > a kernel panic.
> > 
> >     BUG: unable to handle page fault for address: ffff97793ff47000
> >     RIP: ima_restore_measurement_list+0xdc/0x45a
> >     #PF: error_code(0x0000) – not-present page
> > 
> > Other architectures already validate the range with page_is_ram(), as
> > done in commit: cbf9c4b9617b ("of: check previous kernel's
> > ima-kexec-buffer against memory bounds") do a similar check on x86.

Then why isn't there a ima_validate_range() function there which everyone
calls instead of the same check being replicated everywhere?

> > Cc: stable@vger.kernel.org
> > Fixes: b69a2afd5afc ("x86/kexec: Carry forward IMA measurement log on kexec")
> 
> That was via the x86 tree so I assume the x86 team (Boris?) will be
> processing this patch.

Yeah, it is on my to-deal-with-after-the-merge-window pile.

But since you've forced my hand... :-P

> I'll put it into mm.git's mm-hotfixes branch for now, to get a bit of
> testing and to generally track its progress.
> 
> > --- a/arch/x86/kernel/setup.c
> > +++ b/arch/x86/kernel/setup.c
> > @@ -439,9 +439,23 @@ int __init ima_free_kexec_buffer(void)
> >  
> >  int __init ima_get_kexec_buffer(void **addr, size_t *size)
> >  {
> > +	unsigned long start_pfn, end_pfn;
> > +
> >  	if (!ima_kexec_buffer_size)
> >  		return -ENOENT;
> >  
> > +	/*
> > +	 * Calculate the PFNs for the buffer and ensure
> > +	 * they are with in addressable memory.
> 
> "within" ;)
> 
> > +	 */
> > +	start_pfn = PFN_DOWN(ima_kexec_buffer_phys);
> > +	end_pfn = PFN_DOWN(ima_kexec_buffer_phys + ima_kexec_buffer_size - 1);
> > +	if (!pfn_range_is_mapped(start_pfn, end_pfn)) {
> > +		pr_warn("IMA buffer at 0x%llx, size = 0x%zx beyond memory\n",

This error message needs to be made a lot more user-friendly.

And pls do a generic helper as suggested above which ima code calls.

And by looking at the diff, there are two ima_get_kexec_buffer() functions in
the tree which could use some unification too ontop.

Right?

Thx.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Harshit Mogalapalli 1 month, 1 week ago

Hi all,

On 01/12/25 23:49, Borislav Petkov wrote:
> On Mon, Dec 01, 2025 at 09:20:20AM -0800, Andrew Morton wrote:
>> On Wed, 12 Nov 2025 11:30:02 -0800 Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com> wrote:
>>
>>> When the second-stage kernel is booted via kexec with a limiting command
>>> line such as "mem=<size>", the physical range that contains the carried
>>> over IMA measurement list may fall outside the truncated RAM leading to
>>> a kernel panic.
>>>
>>>      BUG: unable to handle page fault for address: ffff97793ff47000
>>>      RIP: ima_restore_measurement_list+0xdc/0x45a
>>>      #PF: error_code(0x0000) – not-present page
>>>
>>> Other architectures already validate the range with page_is_ram(), as
>>> done in commit: cbf9c4b9617b ("of: check previous kernel's
>>> ima-kexec-buffer against memory bounds") do a similar check on x86.
> 
> Then why isn't there a ima_validate_range() function there which everyone
> calls instead of the same check being replicated everywhere?
> 

Thanks for the reviews.

Sure, have tried this, will send a V2 with a generic helper.

>>> Cc: stable@vger.kernel.org
>>> Fixes: b69a2afd5afc ("x86/kexec: Carry forward IMA measurement log on kexec")
>>
>> That was via the x86 tree so I assume the x86 team (Boris?) will be
>> processing this patch.
> 
> Yeah, it is on my to-deal-with-after-the-merge-window pile.
> 
> But since you've forced my hand... :-P
> 
>> I'll put it into mm.git's mm-hotfixes branch for now, to get a bit of
>> testing and to generally track its progress.
>>
>>> --- a/arch/x86/kernel/setup.c
>>> +++ b/arch/x86/kernel/setup.c
>>> @@ -439,9 +439,23 @@ int __init ima_free_kexec_buffer(void)
>>>   
>>>   int __init ima_get_kexec_buffer(void **addr, size_t *size)
>>>   {
>>> +	unsigned long start_pfn, end_pfn;
>>> +
>>>   	if (!ima_kexec_buffer_size)
>>>   		return -ENOENT;
>>>   
>>> +	/*
>>> +	 * Calculate the PFNs for the buffer and ensure
>>> +	 * they are with in addressable memory.
>>
>> "within" ;)
>>

Thanks for spotting.

>>> +	 */
>>> +	start_pfn = PFN_DOWN(ima_kexec_buffer_phys);
>>> +	end_pfn = PFN_DOWN(ima_kexec_buffer_phys + ima_kexec_buffer_size - 1);
>>> +	if (!pfn_range_is_mapped(start_pfn, end_pfn)) {
>>> +		pr_warn("IMA buffer at 0x%llx, size = 0x%zx beyond memory\n",
> 
> This error message needs to be made a lot more user-friendly.
> 
> And pls do a generic helper as suggested above which ima code calls.
> 

Will do, thanks for the suggestion.

> And by looking at the diff, there are two ima_get_kexec_buffer() functions in
> the tree which could use some unification too ontop.
> 

In drivers/of/kexec.c we have:

int __init ima_get_kexec_buffer(void **addr, size_t *size)
{
         int ret, len;
         unsigned long tmp_addr;
         unsigned long start_pfn, end_pfn;
         size_t tmp_size;
         const void *prop;

         prop = of_get_property(of_chosen, "linux,ima-kexec-buffer", &len);
         if (!prop)
                 return -ENOENT;

         ret = do_get_kexec_buffer(prop, len, &tmp_addr, &tmp_size);
         if (ret)
                 return ret;

         /* Do some sanity on the returned size for the ima-kexec buffer */
         if (!tmp_size)
                 return -ENOENT;

         /*
          * Calculate the PFNs for the buffer and ensure
          * they are with in addressable memory.
          */
         start_pfn = PHYS_PFN(tmp_addr);
         end_pfn = PHYS_PFN(tmp_addr + tmp_size - 1);
         if (!page_is_ram(start_pfn) || !page_is_ram(end_pfn)) {
                 pr_warn("IMA buffer at 0x%lx, size = 0x%zx beyond 
memory\n",
                         tmp_addr, tmp_size);
                 return -EINVAL;
         }

         *addr = __va(tmp_addr);
         *size = tmp_size;

         return 0;
}

In arch/x86/kernel/setup.c we have something like:

int __init ima_get_kexec_buffer(void **addr, size_t *size)
{
         if (!ima_kexec_buffer_size)
                 return -ENOENT;

         *addr = __va(ima_kexec_buffer_phys);
         *size = ima_kexec_buffer_size;

         return 0;
}

I will try to generalize common parts in another patch.

Will send a V2 adding ima_validate_range() helper.

Thanks,
Harshit.

> Right?
> 
> Thx.
>

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Jonathan McDowell 2 months, 1 week ago

On Wed, Nov 12, 2025 at 11:30:02AM -0800, Harshit Mogalapalli wrote:
> > 
> When the second-stage kernel is booted via kexec with a limiting command
> line such as "mem=<size>", the physical range that contains the carried
> over IMA measurement list may fall outside the truncated RAM leading to
> a kernel panic.
> 
>     BUG: unable to handle page fault for address: ffff97793ff47000
>     RIP: ima_restore_measurement_list+0xdc/0x45a
>     #PF: error_code(0x0000) – not-present page
> 
> Other architectures already validate the range with page_is_ram(), as
> done in commit: cbf9c4b9617b ("of: check previous kernel's
> ima-kexec-buffer against memory bounds") do a similar check on x86.
> 
> Cc: stable@vger.kernel.org
> Fixes: b69a2afd5afc ("x86/kexec: Carry forward IMA measurement log on kexec")
> Reported-by: Paul Webb <paul.x.webb@oracle.com>
> Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>

Seems legit; this applies on the loaded kernel side, so we do end up
losing the measurements but that matches what OF does.

Reviewed-by: Jonathan McDowell <noodles@meta.com>

> ---
> Have tested the kexec for x86 kernel with IMA_KEXEC enabled and the
> above patch works good. Paul initially reported this on 6.12 kernel but
> I was able to reproduce this on 6.18, so I tried replicating how this
> was fixed in drivers/of/kexec.c
> ---
>  arch/x86/kernel/setup.c | 14 ++++++++++++++
>  1 file changed, 14 insertions(+)
> 
> diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
> index 1b2edd07a3e1..fcef197d180e 100644
> --- a/arch/x86/kernel/setup.c
> +++ b/arch/x86/kernel/setup.c
> @@ -439,9 +439,23 @@ int __init ima_free_kexec_buffer(void)
>  
>  int __init ima_get_kexec_buffer(void **addr, size_t *size)
>  {
> +	unsigned long start_pfn, end_pfn;
> +
>  	if (!ima_kexec_buffer_size)
>  		return -ENOENT;
>  
> +	/*
> +	 * Calculate the PFNs for the buffer and ensure
> +	 * they are with in addressable memory.
> +	 */
> +	start_pfn = PFN_DOWN(ima_kexec_buffer_phys);
> +	end_pfn = PFN_DOWN(ima_kexec_buffer_phys + ima_kexec_buffer_size - 1);
> +	if (!pfn_range_is_mapped(start_pfn, end_pfn)) {
> +		pr_warn("IMA buffer at 0x%llx, size = 0x%zx beyond memory\n",
> +			ima_kexec_buffer_phys, ima_kexec_buffer_size);
> +		return -EINVAL;
> +	}
> +
>  	*addr = __va(ima_kexec_buffer_phys);
>  	*size = ima_kexec_buffer_size;
>  
> -- 
> 2.50.1
> 

-- 
Jonathan McDowell (he/him/his)
Production Engineer | PE Host Integrity
Meta | Facebook UK Ltd

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Harshit Mogalapalli 2 months, 1 week ago

Hi all,

On 13/11/25 01:00, Harshit Mogalapalli wrote:
> When the second-stage kernel is booted via kexec with a limiting command
> line such as "mem=<size>", the physical range that contains the carried
> over IMA measurement list may fall outside the truncated RAM leading to
> a kernel panic.
> 
>      BUG: unable to handle page fault for address: ffff97793ff47000
>      RIP: ima_restore_measurement_list+0xdc/0x45a
>      #PF: error_code(0x0000) – not-present page
> 
> Other architectures already validate the range with page_is_ram(), as
> done in commit: cbf9c4b9617b ("of: check previous kernel's
> ima-kexec-buffer against memory bounds") do a similar check on x86.
> 
> Cc: stable@vger.kernel.org
> Fixes: b69a2afd5afc ("x86/kexec: Carry forward IMA measurement log on kexec")
> Reported-by: Paul Webb <paul.x.webb@oracle.com>
> Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
> ---
> Have tested the kexec for x86 kernel with IMA_KEXEC enabled and the
> above patch works good. Paul initially reported this on 6.12 kernel but
> I was able to reproduce this on 6.18, so I tried replicating how this
> was fixed in drivers/of/kexec.c

ping on this patch.

lore URL: 
https://lore.kernel.org/all/20251112193005.3772542-1-harshit.m.mogalapalli@oracle.com/

Thanks,
Harshit


> ---
>   arch/x86/kernel/setup.c | 14 ++++++++++++++
>   1 file changed, 14 insertions(+)
> 
> diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
> index 1b2edd07a3e1..fcef197d180e 100644
> --- a/arch/x86/kernel/setup.c
> +++ b/arch/x86/kernel/setup.c
> @@ -439,9 +439,23 @@ int __init ima_free_kexec_buffer(void)
>   
>   int __init ima_get_kexec_buffer(void **addr, size_t *size)
>   {
> +	unsigned long start_pfn, end_pfn;
> +
>   	if (!ima_kexec_buffer_size)
>   		return -ENOENT;
>   
> +	/*
> +	 * Calculate the PFNs for the buffer and ensure
> +	 * they are with in addressable memory.
> +	 */
> +	start_pfn = PFN_DOWN(ima_kexec_buffer_phys);
> +	end_pfn = PFN_DOWN(ima_kexec_buffer_phys + ima_kexec_buffer_size - 1);
> +	if (!pfn_range_is_mapped(start_pfn, end_pfn)) {
> +		pr_warn("IMA buffer at 0x%llx, size = 0x%zx beyond memory\n",
> +			ima_kexec_buffer_phys, ima_kexec_buffer_size);
> +		return -EINVAL;
> +	}
> +
>   	*addr = __va(ima_kexec_buffer_phys);
>   	*size = ima_kexec_buffer_size;
>

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Mimi Zohar 2 months, 1 week ago

On Mon, 2025-12-01 at 15:03 +0530, Harshit Mogalapalli wrote:
> Hi all,
> 
> On 13/11/25 01:00, Harshit Mogalapalli wrote:
> > When the second-stage kernel is booted via kexec with a limiting command
> > line such as "mem=<size>", the physical range that contains the carried
> > over IMA measurement list may fall outside the truncated RAM leading to
> > a kernel panic.
> > 
> >      BUG: unable to handle page fault for address: ffff97793ff47000
> >      RIP: ima_restore_measurement_list+0xdc/0x45a
> >      #PF: error_code(0x0000) – not-present page
> > 
> > Other architectures already validate the range with page_is_ram(), as
> > done in commit: cbf9c4b9617b ("of: check previous kernel's
> > ima-kexec-buffer against memory bounds") do a similar check on x86.

It should be obvious that without carrying the measurement list across kexec,
that attestation will fail.  Please mentioned it here in the patch description.

> > 
> > Cc: stable@vger.kernel.org
> > Fixes: b69a2afd5afc ("x86/kexec: Carry forward IMA measurement log on kexec")
> > Reported-by: Paul Webb <paul.x.webb@oracle.com>
> > Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>

Tested-by: Mimi Zohar <zohar@linux.ibm.com>
>

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Ard Biesheuvel 2 months, 1 week ago

On Mon, 1 Dec 2025 at 22:43, Mimi Zohar <zohar@linux.ibm.com> wrote:
>
> On Mon, 2025-12-01 at 15:03 +0530, Harshit Mogalapalli wrote:
> > Hi all,
> >
> > On 13/11/25 01:00, Harshit Mogalapalli wrote:
> > > When the second-stage kernel is booted via kexec with a limiting command
> > > line such as "mem=<size>", the physical range that contains the carried
> > > over IMA measurement list may fall outside the truncated RAM leading to
> > > a kernel panic.
> > >
> > >      BUG: unable to handle page fault for address: ffff97793ff47000
> > >      RIP: ima_restore_measurement_list+0xdc/0x45a
> > >      #PF: error_code(0x0000) – not-present page
> > >
> > > Other architectures already validate the range with page_is_ram(), as
> > > done in commit: cbf9c4b9617b ("of: check previous kernel's
> > > ima-kexec-buffer against memory bounds") do a similar check on x86.
>
> It should be obvious that without carrying the measurement list across kexec,
> that attestation will fail.  Please mentioned it here in the patch description.
>

Couldn't we just use memremap() and be done with it? That will use the
direct map if the memory is mapped, or vmap() it otherwise.

Re: [PATCH] x86/kexec: Add a sanity check on previous kernel's ima kexec buffer

Posted by Mimi Zohar 2 months, 1 week ago

[Cc: Roberto Sassu, linux-integrity]

On Tue, 2025-12-02 at 08:16 +0100, Ard Biesheuvel wrote:
> On Mon, 1 Dec 2025 at 22:43, Mimi Zohar <zohar@linux.ibm.com> wrote:
> > 
> > On Mon, 2025-12-01 at 15:03 +0530, Harshit Mogalapalli wrote:
> > > Hi all,
> > > 
> > > On 13/11/25 01:00, Harshit Mogalapalli wrote:
> > > > When the second-stage kernel is booted via kexec with a limiting command
> > > > line such as "mem=<size>", the physical range that contains the carried
> > > > over IMA measurement list may fall outside the truncated RAM leading to
> > > > a kernel panic.
> > > > 
> > > >      BUG: unable to handle page fault for address: ffff97793ff47000
> > > >      RIP: ima_restore_measurement_list+0xdc/0x45a
> > > >      #PF: error_code(0x0000) – not-present page
> > > > 
> > > > Other architectures already validate the range with page_is_ram(), as
> > > > done in commit: cbf9c4b9617b ("of: check previous kernel's
> > > > ima-kexec-buffer against memory bounds") do a similar check on x86.
> > 
> > It should be obvious that without carrying the measurement list across kexec,
> > that attestation will fail.  Please mention it here in the patch description.
> > 
> 
> Couldn't we just use memremap() and be done with it? That will use the
> direct map if the memory is mapped, or vmap() it otherwise.

No, the IMA measurement list is not a continuous buffer, but a linked list of
records with varying types of fields and field sizes.  The call to
ima_dump_measurement_list() marshals the measurement list into a buffer, while
ima_restore_measurement_list() unmarshals it.

-- 
thanks,

Mimi