On Mon, 2019-02-04 at 14:07 +0200, Jarkko Sakkinen wrote:
> On Mon, Feb 04, 2019 at 10:14:38AM +0100, Roberto Sassu wrote:
> > On 2/1/2019 8:15 PM, Mimi Zohar wrote:
> > > Hi Roberto,
> > >
> > > Sorry for the delayed review. A few comments inline below, minor
> > > suggestions.
> > >
> > > > diff --git a/security/integrity/ima/ima.h b/security/integrity/ima/ima.h
> > > > index cc12f3449a72..e6b2dcb0846a 100644
> > > > --- a/security/integrity/ima/ima.h
> > > > +++ b/security/integrity/ima/ima.h
> > > > @@ -56,6 +56,7 @@ extern int ima_policy_flag;
> > > > extern int ima_hash_algo;
> > > > extern int ima_appraise;
> > > > extern struct tpm_chip *ima_tpm_chip;
> > > > +extern struct tpm_digest *digests;
> > > > /* IMA event related data */
> > > > struct ima_event_data {
> > > > diff --git a/security/integrity/ima/ima_init.c b/security/integrity/ima/ima_init.c
> > > > index 6bb42a9c5e47..296a965b11ef 100644
> > > > --- a/security/integrity/ima/ima_init.c
> > > > +++ b/security/integrity/ima/ima_init.c
> > > > @@ -27,6 +27,7 @@
> > > > /* name for boot aggregate entry */
> > > > static const char boot_aggregate_name[] = "boot_aggregate";
> > > > struct tpm_chip *ima_tpm_chip;
> > > > +struct tpm_digest *digests;
> > >
> > > "digests" is used in the new ima_init_digests() and in
> > > ima_pcr_extend(). It's nice that the initialization routines are
> > > grouped together here in ima_init.c, but wouldn't it better to define
> > > "digests" in ima_queued.c, where it is currently being used?
> > > "digests" could then be defined as static.
> >
> > 'digests' and ima_init_digests() can be moved to ima_queue.c, but I have
> > to add the definition of ima_init_digests() to ima.h. Should I do it?
Yes, I think it is preferable, as it's defined as an __init.
> >
> >
> > > > /* Add the boot aggregate to the IMA measurement list and extend
> > > > * the PCR register.
> > > > @@ -104,6 +105,24 @@ void __init ima_load_x509(void)
> > > > }
> > > > #endif
> > > > +int __init ima_init_digests(void)
> > > > +{
> > > > + int i;
> > > > +
> > > > + if (!ima_tpm_chip)
> > > > + return 0;
> > > > +
> > > > + digests = kcalloc(ima_tpm_chip->nr_allocated_banks, sizeof(*digests),
> > > > + GFP_NOFS);
> > > > + if (!digests)
> > > > + return -ENOMEM;
> > > > +
> > > > + for (i = 0; i < ima_tpm_chip->nr_allocated_banks; i++)
> > > > + digests[i].alg_id = ima_tpm_chip->allocated_banks[i].alg_id;
> > > > +
> > > > + return 0;
> > > > +}
> > > > +
> > > > int __init ima_init(void)
> > > > {
> > > > int rc;
> > > > @@ -125,6 +144,9 @@ int __init ima_init(void)
> > > > ima_load_kexec_buffer();
> > > > + rc = ima_init_digests();
> > >
> > > Ok. Getting the tpm chip is at the beginning of this function.
> > > Deferring allocating "digests" to here, avoids having to free memory
> > > on failure.
> > >
> > > ima_load_kexec_buffer() restores prior measurements, but doesn't
> > > extend the TPM. For anyone reading the code, a short comment above
> > > ima_load_kexec_buffer() would make sense.
> >
> > Ok. Should I resend the last patch again with the fixes you suggested?
>
> Send the full patch set. For me it is easier then to apply the series
> rather than cherry-picking patches from random versions of the patch
> set.
Jarkko, thanks. I've been running with previous versions of this
patchset, and now with this latest version.
Mimi
