linux-kernel - Re: [PATCH 2/3] firmware: qcom_scm: Cleanup code in qcom_scm_assign

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20190723002130.GU30636@minitux>
Date:   Mon, 22 Jul 2019 17:21:30 -0700
From:   Bjorn Andersson <bjorn.andersson@...aro.org>
To:     Stephen Boyd <swboyd@...omium.org>
Cc:     Andy Gross <agross@...nel.org>, linux-kernel@...r.kernel.org,
        linux-arm-msm@...r.kernel.org,
        Ian Jackson <ian.jackson@...rix.com>,
        Julien Grall <julien.grall@....com>,
        Avaneesh Kumar Dwivedi <akdwived@...eaurora.org>
Subject: Re: [PATCH 2/3] firmware: qcom_scm: Cleanup code in
 qcom_scm_assign_mem()

On Mon 22 Jul 17:04 PDT 2019, Stephen Boyd wrote:

> Quoting Bjorn Andersson (2019-07-22 16:27:19)
> > On Fri 17 May 14:09 PDT 2019, Stephen Boyd wrote:
> > 
> > > There are some questionable coding styles in this function. It looks
> > > quite odd to deref a pointer with array indexing that only uses the
> > > first element. Also, destroying an input/output variable halfway through
> > > the function and then overwriting it on success is not clear. It's
> > > better to use a local variable and the kernel macros to step through
> > > each bit set in a bitmask and clearly show where outputs are set.
> > > 
> > > Cc: Ian Jackson <ian.jackson@...rix.com>
> > > Cc: Julien Grall <julien.grall@....com>
> > > Cc: Bjorn Andersson <bjorn.andersson@...aro.org>
> > > Cc: Avaneesh Kumar Dwivedi <akdwived@...eaurora.org>
> > > Signed-off-by: Stephen Boyd <swboyd@...omium.org>
> > > ---
> > >  drivers/firmware/qcom_scm.c | 34 ++++++++++++++++------------------
> > >  include/linux/qcom_scm.h    |  9 +++++----
> > >  2 files changed, 21 insertions(+), 22 deletions(-)
> > > 
> > > diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c
> > > index 0c63495cf269..153f13f72bac 100644
> > > --- a/drivers/firmware/qcom_scm.c
> > > +++ b/drivers/firmware/qcom_scm.c
> > > @@ -443,7 +443,8 @@ EXPORT_SYMBOL(qcom_scm_set_remote_state);
> > >   */
> > >  int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz,
> > >                       unsigned int *srcvm,
> > > -                     struct qcom_scm_vmperm *newvm, int dest_cnt)
> > > +                     const struct qcom_scm_vmperm *newvm,
> > > +                     unsigned int dest_cnt)
> > >  {
> > >       struct qcom_scm_current_perm_info *destvm;
> > >       struct qcom_scm_mem_map_info *mem_to_map;
> > > @@ -458,11 +459,10 @@ int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz,
> > >       int next_vm;
> > >       __le32 *src;
> > >       void *ptr;
> > > -     int ret;
> > > -     int len;
> > > -     int i;
> > > +     int ret, i, b;
> > > +     unsigned long srcvm_bits = *srcvm;
> > >  
> > > -     src_sz = hweight_long(*srcvm) * sizeof(*src);
> > > +     src_sz = hweight_long(srcvm_bits) * sizeof(*src);
> > >       mem_to_map_sz = sizeof(*mem_to_map);
> > >       dest_sz = dest_cnt * sizeof(*destvm);
> > >       ptr_sz = ALIGN(src_sz, SZ_64) + ALIGN(mem_to_map_sz, SZ_64) +
> > > @@ -475,28 +475,26 @@ int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz,
> > >  
> > >       /* Fill source vmid detail */
> > >       src = ptr;
> > > -     len = hweight_long(*srcvm);
> > > -     for (i = 0; i < len; i++) {
> > > -             src[i] = cpu_to_le32(ffs(*srcvm) - 1);
> > > -             *srcvm ^= 1 << (ffs(*srcvm) - 1);
> > > -     }
> > > +     i = 0;
> > > +     for_each_set_bit(b, &srcvm_bits, sizeof(srcvm_bits))
> > 
> > The modem is sad that you only pass 8 here. Changed it to BITS_PER_LONG
> > to include the modem's permission bit and applied all three patches.
> > 
> 
> Ah of course. Thanks.
> 
> BTW, srcvm is an unsigned int, but then we do a bunch of unsigned long
> operations on them. Maybe the whole API should be changed to be more
> explicit about the size of the type, i.e. u64?
> 

It's a bitmap of vmids currently with access to the region and the space
has expanded since I looked at this list time, so the now highest
defined id in the downstream kernel is 42.

So it sounds very reasonable to expand this to a u64.

Regards,
Bjorn