Re: [PATCH] init/main.c: Fix potential static_command_line memory overflow

[Google]

On Fri, 12 Apr 2024 11:51:07 +0800
Yuntao Wang <ytcoode@xxxxxxxxx> wrote:

> On Fri, 12 Apr 2024 00:08:58 +0900, Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx> wrote:
> 
> > On Thu, 11 Apr 2024 22:27:23 +0800
> > Yuntao Wang <ytcoode@xxxxxxxxx> wrote:
> > 
> > > On Thu, 11 Apr 2024 16:21:37 +0300, Mike Rapoport <rppt@xxxxxxxxxx> wrote:
> > > > On Thu, Apr 11, 2024 at 09:23:47AM +0200, Geert Uytterhoeven wrote:
> > > > > CC Hiramatsu-san
> > > > > 
> > > > > On Thu, Apr 11, 2024 at 5:25 AM Yuntao Wang <ytcoode@xxxxxxxxx> wrote:
> > > > > > We allocate memory of size 'xlen + strlen(boot_command_line) + 1' for
> > > > > > static_command_line, but the strings copied into static_command_line are
> > > > > > extra_command_line and command_line, rather than extra_command_line and
> > > > > > boot_command_line.
> > > > > >
> > > > > > When strlen(command_line) > strlen(boot_command_line), static_command_line
> > > > > > will overflow.
> > > > 
> > > > Can this ever happen? 
> > > > Did you observe the overflow or is this a theoretical bug?
> > > 
> > > I didn't observe the overflow, it's just a theoretical bug.
> > > 
> > > > > > Fixes: f5c7310ac73e ("init/main: add checks for the return value of memblock_alloc*()")
> > > > 
> > > > f5c7310ac73e didn't have the logic for calculating allocation size, we
> > > > surely don't want to go back that far wiht Fixes.
> > > 
> > > Before commit f5c7310ac73e, the memory size allocated for static_command_line
> > > was 'strlen(command_line) + 1', but commit f5c7310ac73e changed this size
> > > to 'strlen(boot_command_line) + 1'. I think this should be wrong.
> > 
> > Ah, OK. that sounds reasonable. 
> > 
> > > 
> > > > > > Signed-off-by: Yuntao Wang <ytcoode@xxxxxxxxx>
> > > > > > ---
> > > > > >  init/main.c | 8 +++++---
> > > > > >  1 file changed, 5 insertions(+), 3 deletions(-)
> > > > > >
> > > > > > diff --git a/init/main.c b/init/main.c
> > > > > > index 2ca52474d0c3..a7b1f5f3e3b6 100644
> > > > > > --- a/init/main.c
> > > > > > +++ b/init/main.c
> > > > > > @@ -625,11 +625,13 @@ static void __init setup_command_line(char *command_line)
> > > > > >         if (extra_init_args)
> > > > > >                 ilen = strlen(extra_init_args) + 4; /* for " -- " */
> > > > > >
> > > > > > -       len = xlen + strlen(boot_command_line) + 1;
> > > > > > +       len = xlen + strlen(boot_command_line) + ilen + 1;
> > > > > >
> > > > > > -       saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
> > > > > > +       saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
> > > > > >         if (!saved_command_line)
> > > > > > -               panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
> > > > > > +               panic("%s: Failed to allocate %zu bytes\n", __func__, len);
> > > > > > +
> > > > > > +       len = xlen + strlen(command_line) + 1;
> > 
> > Ah, I missed this line. Sorry. So this looks good to me but you don't need any
> > other lines, because those are not related to the bug you want to fix.
> > Please just focus on 1 fix.
> 
> Hi Masami,
> 
> Do I need to split this patch into two? Or should I just repost this patch
> with any other lines not related to this bug removed?

Latter one should be easier. Only add above one line and just explain that this
recovers strlen(command_line) which was miss-consolidated with strlen(boot_command_line)
in the commit f5c7310ac73e ("init/main: add checks for the return value of
memblock_alloc*()"). Simple fix does not confuse reviewers.

> 
> Actually, I think these lines are still necessary as they make the code
> look a bit cleaner.

That is a cleanup, and should be separated from bugfix, because bugfix must
be backported but the cleanup doesn't. As far as I can see, the cleanup
part can not apply to the commit f5c7310ac73e.

Thank you,

> 
> Thanks,
> Yuntao
> 
> > Thank you,
> > 
> > > > > >
> > > > > >         static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
> > > > > >         if (!static_command_line)
> > > > > 
> > > > > Gr{oetje,eeting}s,
> > > > > 
> > > > >                         Geert
> > > > > 
> > > > > -- 
> > > > > Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
> > > > > 
> > > > > In personal conversations with technical people, I call myself a hacker. But
> > > > > when I'm talking to journalists I just say "programmer" or something like that.
> > > > >                                 -- Linus Torvalds
> > > > 
> > > > -- 
> > > > Sincerely yours,
> > > > Mike.
> > 
> > 
> > -- 
> > Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>


-- 
Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>