Re: [PATCH v2 3/4][next] scsi: megaraid_sas: Replace one-element array with flexible-array member in MR_DRV_RAID_MAP

["Gustavo A. R. Silva" <gustavoars@kernel.org>]

On Thu, Jun 23, 2022 at 08:19:49AM -0700, Kees Cook wrote:
> On Thu, Jun 23, 2022 at 05:14:01AM +0200, Gustavo A. R. Silva wrote:
> > On Thu, Jun 23, 2022 at 03:45:33AM +0200, Gustavo A. R. Silva wrote:
> > > On Wed, Jun 22, 2022 at 03:26:38PM -0700, Kees Cook wrote:
> > > > On Wed, Aug 04, 2021 at 11:20:04PM -0500, Gustavo A. R. Silva wrote:
> > > > > Replace one-element array with a flexible-array member in struct
> > > > > MR_DRV_RAID_MAP and use the flex_array_size() helper.
> > > > > 
> > > > > This helps with the ongoing efforts to globally enable -Warray-bounds
> > > > > and get us closer to being able to tighten the FORTIFY_SOURCE routines
> > > > > on memcpy().
> > > > > 
> > > > > Link: https://en.wikipedia.org/wiki/Flexible_array_member
> > > > > Link: https://www.kernel.org/doc/html/v5.10/process/deprecated.html#zero-length-and-one-element-arrays
> > > > > Link: https://github.com/KSPP/linux/issues/79
> > > > > Link: https://github.com/KSPP/linux/issues/109
> > > > > Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
> > > > 
> > > > I'd really like to see this fixed. :) I'm running into this 1-element
> > > > array problem now with UBSAN_BOUNDS:
> > > 
> > > Wow; another forgoten patch from the times we didn't have Patchwork. :) 
> > > 
> > > > 
> > > > [   10.011173] UBSAN: array-index-out-of-bounds in /build/linux-WLUive/linux-5.15.0/drivers/scsi/megaraid/megaraid_sas_fp.c:103:32
> > > > [   10.087824] index 1 is out of range for type 'MR_LD_SPAN_MAP [1]'
> > > > 
> > > > and I'm not the only one:
> > > > 
> > > > https://bugzilla.kernel.org/show_bug.cgi?id=215943
> > > 
> > > It's actually great that other people are running into these issues now.
> > > That could only means that we should fixed ASAP. :)
> > > 
> > > We also have this other series that hasn't been applied yet:
> > > 
> > > https://lore.kernel.org/linux-hardening/cover.1645513670.git.gustavoars@kernel.org/
> > > 
> > > > 
> > > > > ---
> > > > > Changes in v2:
> > > > >  - None.
> > > > > 
> > > > >  drivers/scsi/megaraid/megaraid_sas_fp.c     | 6 +++---
> > > > >  drivers/scsi/megaraid/megaraid_sas_fusion.h | 2 +-
> > > > >  2 files changed, 4 insertions(+), 4 deletions(-)
> > > > > 
> > > > > diff --git a/drivers/scsi/megaraid/megaraid_sas_fp.c b/drivers/scsi/megaraid/megaraid_sas_fp.c
> > > > > index da1cad1ee123..9cb36ef96c2c 100644
> > > > > --- a/drivers/scsi/megaraid/megaraid_sas_fp.c
> > > > > +++ b/drivers/scsi/megaraid/megaraid_sas_fp.c
> > > > > @@ -229,8 +229,8 @@ static int MR_PopulateDrvRaidMap(struct megasas_instance *instance, u64 map_id)
> > > > >  					le32_to_cpu(desc_table->raid_map_desc_offset));
> > > > >  				memcpy(pDrvRaidMap->ldSpanMap,
> > > > >  				       fw_map_dyn->ld_span_map,
> > > > > -				       sizeof(struct MR_LD_SPAN_MAP) *
> > > > > -				       le32_to_cpu(desc_table->raid_map_desc_elements));
> > > > > +				       flex_array_size(pDrvRaidMap, ldSpanMap,
> > > > > +				       le32_to_cpu(desc_table->raid_map_desc_elements)));
> > > > >  			break;
> > > > >  			default:
> > > > >  				dev_dbg(&instance->pdev->dev, "wrong number of desctableElements %d\n",
> > > > > @@ -254,7 +254,7 @@ static int MR_PopulateDrvRaidMap(struct megasas_instance *instance, u64 map_id)
> > > > >  			pDrvRaidMap->ldTgtIdToLd[i] =
> > > > >  				(u16)fw_map_ext->ldTgtIdToLd[i];
> > > > >  		memcpy(pDrvRaidMap->ldSpanMap, fw_map_ext->ldSpanMap,
> > > > > -		       sizeof(struct MR_LD_SPAN_MAP) * ld_count);
> > > > > +		       flex_array_size(pDrvRaidMap, ldSpanMap, ld_count));
> > > > >  		memcpy(pDrvRaidMap->arMapInfo, fw_map_ext->arMapInfo,
> > > > >  		       sizeof(struct MR_ARRAY_INFO) * MAX_API_ARRAYS_EXT);
> > > > >  		memcpy(pDrvRaidMap->devHndlInfo, fw_map_ext->devHndlInfo,
> > > > > diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.h b/drivers/scsi/megaraid/megaraid_sas_fusion.h
> > > > > index 9adb8b30f422..5fe2f7a6eebe 100644
> > > > > --- a/drivers/scsi/megaraid/megaraid_sas_fusion.h
> > > > > +++ b/drivers/scsi/megaraid/megaraid_sas_fusion.h
> > > > > @@ -1182,7 +1182,7 @@ struct MR_DRV_RAID_MAP {
> > > > >  		devHndlInfo[MAX_RAIDMAP_PHYSICAL_DEVICES_DYN];
> > > > >  	u16 ldTgtIdToLd[MAX_LOGICAL_DRIVES_DYN];
> > > > >  	struct MR_ARRAY_INFO arMapInfo[MAX_API_ARRAYS_DYN];
> > > > > -	struct MR_LD_SPAN_MAP      ldSpanMap[1];
> > > > > +	struct MR_LD_SPAN_MAP      ldSpanMap[];
> > > > >  
> > > > >  };
> > > > >  
> > > > 
> > > > I think this patch is incomplete, and the wrapping struct needs to be
> > > > adjusted too:
> > > > 
> > > > @@ -1193,7 +1193,7 @@ struct MR_DRV_RAID_MAP {
> > > >  struct MR_DRV_RAID_MAP_ALL {
> > > >  
> > > >         struct MR_DRV_RAID_MAP raidMap;
> > > > -       struct MR_LD_SPAN_MAP ldSpanMap[MAX_LOGICAL_DRIVES_DYN - 1];
> > > > +       struct MR_LD_SPAN_MAP ldSpanMap[MAX_LOGICAL_DRIVES_DYN];
> > > >  } __packed;
> > 
> > BTW, I'd really like to get some input from the maintainers of this
> > code. :)
> 
> Agreed, though if we get 0 output from a diffoscope of the object files,
> I think it's safe to carry such patches in our trees. This is how I've
> been testing:

Which object files are you comparing here? because I don't see the zero
change when comparing the before and after of megaraid_sas_fp.o with
the change you propose.

> 
> $ make allmodconfig
> $ ./scripts/config -d GCOV_KERNEL -d KCOV -d GCC_PLUGINS -d IKHEADERS -d KASAN -d UBSAN
> $ make olddefconfig
> $ make the/path/to/code.o
> $ cp the/path/to/code.o the/path/to/code.before
> $ *apply patch*
> $ make the/path/to/code.o
> $ cp the/path/to/code.o the/path/to/code.after
> $ diffoscope the/path/to/code.before the/path/to/code.after

Yep; that's how we do it. However, I think the zero change is the
exception, not the norm in these sorts of changes.

The machine code could easily change after refactoring the code a
bit as a consequence of the flex-array transformation. It's not like
we are merely removing the 1 from the array declaration. We are also
making use of helpers like struct_size(), flex_array_size() and in
some cases we are making the code to stop allocating some too-many
extra bytes (which usually account for the size of one item of the
trailing array) that are never actually used. I think that could
easily impact how the compiler ultimately arrange the code.

--
Gustavo