Re: [dpdk-dev] [PATCH] lib/cryptodev: fix driver name comparison

["Trahe, Fiona" <fiona.trahe@intel.com>]

Hi Anoob

> -----Original Message-----
> From: Anoob Joseph [mailto:anoobj@marvell.com]
> Sent: Saturday, February 23, 2019 6:12 AM
> To: Trahe, Fiona <fiona.trahe@intel.com>; Akhil Goyal <akhil.goyal@nxp.com>; Doherty, Declan
> <declan.doherty@intel.com>; De Lara Guarch, Pablo <pablo.de.lara.guarch@intel.com>; Yigit, Ferruh
> <ferruh.yigit@intel.com>; Thomas Monjalon <thomas@monjalon.net>
> Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Narayana Prasad Raju Athreya
> <pathreya@marvell.com>; dev@dpdk.org; Ankur Dwivedi <adwivedi@marvell.com>
> Subject: RE: [PATCH] lib/cryptodev: fix driver name comparison
> 
> Hi Fiona,
> 
> > -----Original Message-----
> > From: Trahe, Fiona <fiona.trahe@intel.com>
> > Sent: Friday, February 22, 2019 9:09 PM
> > To: Anoob Joseph <anoobj@marvell.com>; Akhil Goyal <akhil.goyal@nxp.com>;
> > Doherty, Declan <declan.doherty@intel.com>; De Lara Guarch, Pablo
> > <pablo.de.lara.guarch@intel.com>
> > Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Narayana Prasad Raju
> > Athreya <pathreya@marvell.com>; dev@dpdk.org; Ankur Dwivedi
> > <adwivedi@marvell.com>; Trahe, Fiona <fiona.trahe@intel.com>
> > Subject: RE: [PATCH] lib/cryptodev: fix driver name comparison
> >
> > Hi Anoob,
> >
> > > > > > @@ -542,8 +543,8 @@ rte_cryptodev_get_dev_id(const char *name)
> > > > > >  		return -1;
> > > > > >
> > > > > >  	for (i = 0; i < cryptodev_globals.nb_devs; i++)
> > > > > > -		if ((strcmp(cryptodev_globals.devs[i].data->name, name)
> > > > > > -				== 0) &&
> > > > > > +		if ((strncmp(cryptodev_globals.devs[i].data->name,
> > name,
> > > > > > +				RTE_CRYPTODEV_NAME_MAX_LEN)
> > == 0)
> > > > &&
> > > > [Fiona] Is this safe? The const passed to this may not be the full
> > > > length of RTE_CRYPTODEV_NAME_MAX_LEN. Does this prototype need to
> > > > specify that a full length const filled with trailing zeros must be
> > > > passed in? And if so is this an ABI breakage?
> > > >
> > >
> > > [Anoob] strcmp itself is not safe when we have buffers which are not
> > > NULL terminated. Strncmp will make sure the check won't exceed
> > RTE_CRYPTODEV_NAME_MAX_LEN.
> > >
> > > From man page, "The strncmp() function is similar, except it only
> > > compares the first (at most) n bytes of
> > > s1 and s2."
> > >
> > > The main issue here is the usage of strncmp with strlen(driver_name),
> > > as in the below cases. Strlen will return string length, which doesn't
> > > include \0. strcmp is good enough to fix the issue. But usage of
> > > strcmp would assume that the const is filled with trailing zero. IMO, none of
> > these options are really safe. So please advise on what would be the best
> > solution here. I'll revise the patch accordingly.
> > [Fiona] I agree and think it is safest as you've coded it. However I'd suggest
> > adding a comment on the relevant APIs saying that the string must be passed in
> > in a buffer of size <use relevant #define> with trailing zeros.
> 
> [Anoob] Do you want this patch to address that? And wouldn't specifying something like that explicitly, be
> an ABI breakage?
[Fiona] Yes, I think it should be in this patch as this patch is causing it.
But it's up to the maintainers what's acceptable - it seems to me that it's an ABI breakage, avoiding
saying it explicitly doesn't make it less so.
 
> 
> Also, I think the same is applicable for other similar functions (rte_eth_dev_get_port_by_name() etc),
> wherever we expect a string. Please do share your thoughts on what all I should include in this patch.
> 
> Thanks,
> Anoob