linux-kernel - Re: [PATCH v2] x86/vdso: Change return type to vm_fault

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: <CAFqt6zZ+dp9gvF3kx2OO6kv_GH0Nrm2e31Om_c9y78Mu+Krzfw@mail.gmail.com>
Date:   Tue, 3 Jul 2018 20:46:02 +0530
From:   Souptick Joarder <jrdr.linux@...il.com>
To:     Matthew Wilcox <willy@...radead.org>
Cc:     Andy Lutomirski <luto@...capital.net>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        "H. Peter Anvin" <hpa@...or.com>, X86 ML <x86@...nel.org>,
        linux-kernel@...r.kernel.org, brajeswar.linux@...il.com,
        Sabyasachi Gupta <sabyasachi.linux@...il.com>
Subject: Re: [PATCH v2] x86/vdso: Change return type to vm_fault_t for fault handlers

On Tue, Jul 3, 2018 at 5:41 PM, Matthew Wilcox <willy@...radead.org> wrote:
> On Tue, Jul 03, 2018 at 05:18:30PM +0530, Souptick Joarder wrote:
>> On Tue, Jul 3, 2018 at 4:47 PM, Matthew Wilcox <willy@...radead.org> wrote:
>> > On Mon, Jun 25, 2018 at 11:27:37PM +0530, Souptick Joarder wrote:
>> >> Use new return type vm_fault_t for both fault handler
>> >> vdso_fault() and vvar_fault(). Previously vm_insert_pfn()
>> >> returns err which has to mapped into VM_FAULT_* type.
>> >> The new function vmf_insert_pfn() will replace this
>> >> inefficiency by returning correct VM_FAULT_* type.
>> >
>> >> @@ -105,10 +105,10 @@ static int vvar_fault(const struct vm_special_mapping *sm,
>> >>        * the page past the end of the vvar mapping.
>> >>        */
>> >>       if (sym_offset == 0)
>> >> -             return VM_FAULT_SIGBUS;
>> >> +             return ret;
>> >>
>> >>       if (sym_offset == image->sym_vvar_page) {
>> >> -             ret = vm_insert_pfn(vma, vmf->address,
>> >> +             ret = vmf_insert_pfn(vma, vmf->address,
>> >>                                   __pa_symbol(&__vvar_page) >> PAGE_SHIFT);
>> >>       } else if (sym_offset == image->sym_pvclock_page) {
>> >>               struct pvclock_vsyscall_time_info *pvti =
>> >> @@ -124,14 +124,11 @@ static int vvar_fault(const struct vm_special_mapping *sm,
>> >
>> > Haven't you missed converting vm_insert_pfn_prot() at line 117?
>> > Did you test-compile this?
>> >
>>
>> I left it intentionally in this patch. When we will be replacing
>> vm_insert_foo() with new API vmf_insert_foo(), vm_insert_pfn_prot
>> need to be changed to return vm_fault_t type.
>>
>> I will change the return type of vm_insert_pfn_prot() in that patch.
>
> What?!  That's just broken.  vm_insert_pfn_prot returns an errno, so
> this patch introduces a bug where sometimes 'ret' contains a vm_fault_t
> and sometimes contains an errno.  That's exactly the kind of thing this
> patch series is supposed to be *preventing*, not introducing!
>

It was my wrong understanding which create this issue. Sorry about it.
I though to change the return type of vm_insert_pfn_prot() to vm_fault_t
type directly along with the caller of vm_insert_pfn_prot(), but in a separate
patch.

But introducing vmf_insert_pfn_prot() is much simpler approach here.

Sorry for making things complex.