Re: [PATCH] firmware: make sure the fw file size is not 0

[Luciano Coelho]

On Wed, 2013-01-16 at 11:43 +0800, Ming Lei wrote:
> On Tue, Jan 15, 2013 at 4:43 PM, Luciano Coelho <coelho@xxxxxx> wrote:
> > If the requested firmware file size is 0 bytes in the filesytem, we
> > will try to vmalloc(0), which causes a warning:
> >
> > [37834.750274] vmalloc: allocation failure: 0 bytes
> > [37834.756347] kworker/1:1: page allocation failure: order:0, mode:0xd2
> > [37834.763214] [<c001ec60>] (unwind_backtrace+0x0/0x13c) from [<c07171ac>] (dump_stack+0x20/0x24)
> > [37834.772430] [<c07171ac>] (dump_stack+0x20/0x24) from [<c010c7ac>] (warn_alloc_failed+0xc8/0x120)
> > [37834.781799] [<c010c7ac>] (warn_alloc_failed+0xc8/0x120) from [<c0142860>] (__vmalloc_node_range+0x16)
> > [37834.792663] [<c0142860>] (__vmalloc_node_range+0x164/0x208) from [<c0142950>] (__vmalloc_node+0x4c/0)
> > [37834.802703] [<c0142950>] (__vmalloc_node+0x4c/0x58) from [<c0142994>] (vmalloc+0x38/0x44)
> > [37834.811401] [<c0142994>] (vmalloc+0x38/0x44) from [<c03f729c>] (_request_firmware_load+0x220/0x6b0)
> > [37834.820983] [<c03f729c>] (_request_firmware_load+0x220/0x6b0) from [<c03f7f68>] (request_firmware+0x)
> > [37834.831451] [<c03f7f68>] (request_firmware+0x64/0xc8) from [<bf51b9a8>] (wl18xx_setup+0xb4/0x570 [wl)
> > [37834.841918] [<bf51b9a8>] (wl18xx_setup+0xb4/0x570 [wl18xx]) from [<bf4dfdac>] (wlcore_nvs_cb+0x64/0x)
> > [37834.852844] [<bf4dfdac>] (wlcore_nvs_cb+0x64/0x9f8 [wlcore]) from [<c03f7e98>] (request_firmware_wor)
> > [37834.864227] [<c03f7e98>] (request_firmware_work_func+0x94/0x100) from [<c0072d7c>] (process_one_work)
> > [37834.875122] [<c0072d7c>] (process_one_work+0x1d0/0x750) from [<c00736f8>] (worker_thread+0x184/0x4ac)
> > [37834.884918] [<c00736f8>] (worker_thread+0x184/0x4ac) from [<c007a2d4>] (kthread+0xb4/0xc0)
> > [37834.893707] [<c007a2d4>] (kthread+0xb4/0xc0) from [<c0015490>] (ret_from_fork+0x14/0x20)
> >
> > To fix this, check whether the file size is less than or equal to zero
> > in fw_read_file_contents().
> >
> > Cc: stable <stable@xxxxxxxxxxxxxxx> [3.7]
> > Signed-off-by: Luciano Coelho <coelho@xxxxxx>
> > ---
> >  drivers/base/firmware_class.c |    2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/base/firmware_class.c b/drivers/base/firmware_class.c
> > index d814603..b392b35 100644
> > --- a/drivers/base/firmware_class.c
> > +++ b/drivers/base/firmware_class.c
> > @@ -305,7 +305,7 @@ static bool fw_read_file_contents(struct file *file, struct firmware_buf *fw_buf
> >         char *buf;
> >
> >         size = fw_file_size(file);
> > -       if (size < 0)
> > +       if (size <= 0)
> >                 return false;
> 
> One trivial thing is that if we should return false on zero size firmware?
> 
> If we think zero size firmware file as a good fw image, we should handle
> the case and return true. Otherwise, false should be returned. But that is
> not a big deal, so looks fine for me.

That's a good point.  It is what I was trying out when I found this bug.

With our chip (WiLink) we use a binary containing "patches" to modify
the actual firmware that sits in the module's ROM.  If there is nothing
to be modified, it would make sense to have a zero-sized binary.  This
way we could make sure that applying no "patches" is done explicitly
(and not because of a misplaced or missing firmware).

In any case, as far as I can see, udev doesn't accept zero-sized
firmwares either [1], so it makes sense to do the same in the kernel.

[1] As of version 197:
http://cgit.freedesktop.org/systemd/systemd/tree/src/udev/udev-builtin-firmware.c?id=v197#n142


--
Luca.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/