linux-kernel - Re: [PATCH 2/4] KVM: selftests: Add helper macros for ioctl()s that return file descriptors

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

Message-ID: <ZM1EkRMf23e2YUBs@google.com>
Date:   Fri, 4 Aug 2023 11:33:53 -0700
From:   Sean Christopherson <seanjc@...gle.com>
To:     Colton Lewis <coltonlewis@...gle.com>
Cc:     Oliver Upton <oliver.upton@...ux.dev>, maz@...nel.org,
        linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.linux.dev,
        linux-kernel@...r.kernel.org, mhal@...x.co
Subject: Re: [PATCH 2/4] KVM: selftests: Add helper macros for ioctl()s that
 return file descriptors

On Fri, Aug 04, 2023, Colton Lewis wrote:
> Oliver Upton <oliver.upton@...ux.dev> writes:
> 
> > Hi Sean,
> 
> > On Thu, Aug 03, 2023 at 05:42:24PM -0700, Sean Christopherson wrote:
> > > Add KVM, VM, and vCPU scoped helpers for ioctl()s that return file
> > > descriptors, i.e. deduplicate code for asserting success on ioctls() for
> > > which a positive return value, not just zero, is considered success.
> 
> > > Signed-off-by: Sean Christopherson <seanjc@...gle.com>
> 
> > I appreciate the desire to eliminate duplicate code, but I think the
> > naming just muddies the waters. TBH, when I first read the diff w/o the
> > changelog, I thought you were describing the input fd (i.e. 'kvm_fd',
> > 'vm_fd', 'vcpu_fd'). I don't think explicitly spelling out the condition
> > each time (i.e. ret >= 0) is all that difficult.
> 
> Couldn't ret >= 0 be the assert condition for everything? Don't see why
> there needs to be different helpers to check == 0 and >= 0.
> 
> Unless I'm missing something, error returns are only ever negative.

Using "ret >= 0" would work in the sense that the tests wouldn't fail, but it
would degrade our test coverage, e.g. selftests wouldn't detect KVM bugs where
an ioctl() unexpectedly returns a non-zero, positive value.

The other wrinkle is that selftests need to actually consume the return value for
ioctl()s that return a positive value, i.e. the fd (or whatever it is) needs to
propagated up the stack.  I.e. all of the generic ioctl() macros would need to
"return" the value, which I really don't want to do because that would (re)open
the gates for having helpers that return an int, even though the only possible
return value is '0'.