Re: [PATCH v11 2/9] ima: define and call ima_alloc_kexec_file_buf()

[Mimi Zohar]

On Tue, 2025-04-08 at 12:39 +0800, Baoquan He wrote:
> On 04/08/25 at 12:07am, Mimi Zohar wrote:
> > On Wed, 2025-04-02 at 05:47 -0700, steven chen wrote:
> > > In the current implementation, the ima_dump_measurement_list() API is 
> > > called during the kexec "load" phase, where a buffer is allocated and 
> > > the measurement records are copied. Due to this, new events added after
> > > kexec load but before kexec execute are not carried over to the new kernel
> > > during kexec operation
> > 
> > Repeating this here is unnecessary.
> > > 
> > > To allow the buffer allocation and population to be separated into distinct
> > > steps, make the function local seq_file "ima_kexec_file" to a file variable.
> > 
> > This change was already made in [PATCH v11 1/9] ima: rename variable the
> > set_file "file" to "ima_kexec_file".  Please remove.
> > 
> > > 
> > > Carrying the IMA measurement list across kexec requires allocating a
> > > buffer and copying the measurement records.  Separate allocating the
> > > buffer and copying the measurement records into separate functions in
> > > order to allocate the buffer at kexec 'load' and copy the measurements
> > > at kexec 'execute'.
> > > 
> > > Signed-off-by: Tushar Sugandhi <tusharsu@xxxxxxxxxxxxxxxxxxx>
> > > Signed-off-by: steven chen <chenste@xxxxxxxxxxxxxxxxxxx>
> > > ---
> > >  security/integrity/ima/ima_kexec.c | 46 +++++++++++++++++++++++-------
> > >  1 file changed, 35 insertions(+), 11 deletions(-)
> > > 
> > > diff --git a/security/integrity/ima/ima_kexec.c b/security/integrity/ima/ima_kexec.c
> > > index 650beb74346c..b12ac3619b8f 100644
> > > --- a/security/integrity/ima/ima_kexec.c
> > > +++ b/security/integrity/ima/ima_kexec.c
> > > @@ -15,26 +15,46 @@
> > >  #include "ima.h"
> > >  
> > >  #ifdef CONFIG_IMA_KEXEC
> > > +static struct seq_file ima_kexec_file;
> > > +
> > > +static void ima_free_kexec_file_buf(struct seq_file *sf)
> > > +{
> > > +	vfree(sf->buf);
> > > +	sf->buf = NULL;
> > > +	sf->size = 0;
> > > +	sf->read_pos = 0;
> > > +	sf->count = 0;
> > > +}
> > > +
> > > +static int ima_alloc_kexec_file_buf(size_t segment_size)
> > > +{
> > > +	ima_free_kexec_file_buf(&ima_kexec_file);
> > 
> > After moving the vfree() here at this stage in the patch set, the IMA
> > measurement list fails to verify when doing two consecutive "kexec -s -l"
> > with/without a "kexec -s -u" in between.  Only after "ima: kexec: move IMA log
> > copy from kexec load to execute" the IMA measurement list verifies properly with
> > the vfree() here.
> 
> I also noticed this, patch 7 will remedy this. Put patch 7 just after
> this patch or squash it into this patch?
> 
> [PATCH v11 7/9] ima: verify if the segment size has changed

I'm glad you noticed this too!  I've been staring at it for a while, not knowing
what to do.

"ima: verify if the segment size has changed" is new to v11.  It was originally
part of this patch.  My comment on v10 was:

The call to ima_reset_kexec_file() in ima_add_kexec_buffer() resets
ima_kexec_file.buf() hiding the fact that the above test always fails and falls
through.  As a result, 'buf' is always being re-allocated.

and

Instead of adding and then removing the ima_reset_kexec_file() call from
ima_add_kexec_buffer(), defer adding the segment size test to when it is
actually possible for the segment size to change. Please make the segment size
test as a separate patch.

ima_reset_kexec_file() will then only be called by ima_free_kexec_file_buf().
Inline the ima_reset_kexec_file() code in ima_free_kexec_file_buf().

> 
> > 
> > > +
> > > +	/* segment size can't change between kexec load and execute */
> > > +	ima_kexec_file.buf = vmalloc(segment_size);
> > > +	if (!ima_kexec_file.buf)
> > > +		return -ENOMEM;
> > > +
> > > +	ima_kexec_file.size = segment_size;
> > > +	ima_kexec_file.read_pos = 0;
> > > +	ima_kexec_file.count = sizeof(struct ima_kexec_hdr);	/* reserved space */
> > > +
> > > +	return 0;
> > > +}
> > > +
> > 
> 
>