Re: [PATCH] Fix memory leak in vc_resize/vc_allocate

["Catalin Marinas" <catalin.marinas@gmail.com>]

On 14/08/06, Andrew Morton <akpm@osdl.org> wrote:
> On Thu, 10 Aug 2006 15:22:21 +0100
> Catalin Marinas <catalin.marinas@arm.com> wrote:
> > diff --git a/drivers/char/vt.c b/drivers/char/vt.c
> > index da7e66a..31c8b32 100644
> > --- a/drivers/char/vt.c
> > +++ b/drivers/char/vt.c
> > @@ -730,7 +730,8 @@ int vc_allocate(unsigned int currcons)    /
> >           visual_init(vc, currcons, 1);
> >           if (!*vc->vc_uni_pagedir_loc)
> >               con_set_default_unimap(vc);
> > -         vc->vc_screenbuf = kmalloc(vc->vc_screenbuf_size, GFP_KERNEL);
> > +         if (!vc->vc_kmalloced)
> > +             vc->vc_screenbuf = kmalloc(vc->vc_screenbuf_size, GFP_KERNEL);
> >           if (!vc->vc_screenbuf) {
> >               kfree(vc);
> >               vc_cons[currcons].d = NULL;
>
> hm.  Maybe.  I'd worry that the memory at vc->vc_screenbuf isn't of the
> correct size and this patch will convert a leak into a buffer overrun.

My initial attempt was to free the kmalloc'ed buffer and reallocate it
in vc_allocate (similar to the code in vc_resize) but I realised that
vc_screenbuf kmalloced block is always of the vs_screenbuf_size
length.

> Also, what's up with this, in vc_resize()?
>
>         if (vc->vc_kmalloced)
>                 kfree(vc->vc_screenbuf);
>         vc->vc_screenbuf = newscreen;
>         vc->vc_kmalloced = 1;
>
> if vc->vc_kmalloced means "there is kmalloced memory at vc->vc_screenbuf"
> then this is wrong.

I think the above should be fine because vc_resize can be called
either via vc_allocate->visual_init->fbcon_init when the buffer was
not yet allocated or via a tty ioctl etc. and it needs to free the old
buffer.

-- 
Catalin