[PATCH] mtd: jedec_probe: Fix crash in jedec_read_mfr()

[Boris Brezillon]

On Tue, 27 Mar 2018 21:02:08 +0200
Richard Weinberger <richard at nod.at> wrote:

> Am Samstag, 3. März 2018, 23:29:03 CEST schrieb Linus Walleij:
> > It turns out that the loop where we read manufacturer
> > jedec_read_mfd() can under some circumstances get a
> > CFI_MFR_CONTINUATION repeatedly, making the loop go
> > over all banks and eventually hit the end of the
> > map and crash because of an access violation:
> > 
> > Unable to handle kernel paging request at virtual address c4980000
> > pgd = (ptrval)
> > [c4980000] *pgd=03808811, *pte=00000000, *ppte=00000000
> > Internal error: Oops: 7 [#1] PREEMPT ARM
> > CPU: 0 PID: 1 Comm: swapper Not tainted 4.16.0-rc1+ #150
> > Hardware name: Gemini (Device Tree)
> > PC is at jedec_probe_chip+0x6ec/0xcd0
> > LR is at 0x4
> > pc : [<c03a2bf4>]    lr : [<00000004>]    psr: 60000013
> > sp : c382dd18  ip : 0000ffff  fp : 00000000
> > r10: c0626388  r9 : 00020000  r8 : c0626340
> > r7 : 00000000  r6 : 00000001  r5 : c3a71afc  r4 : c382dd70
> > r3 : 00000001  r2 : c4900000  r1 : 00000002  r0 : 00080000
> > Flags: nZCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment none
> > Control: 0000397f  Table: 00004000  DAC: 00000053
> > Process swapper (pid: 1, stack limit = 0x(ptrval))
> > 
> > Fix this by breaking the loop with a return 0 if
> > the offset exceeds the map size.
> > 
> > Signed-off-by: Linus Walleij <linus.walleij at linaro.org>
> > ---
> >  drivers/mtd/chips/jedec_probe.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/drivers/mtd/chips/jedec_probe.c
> > b/drivers/mtd/chips/jedec_probe.c index 7c0b27d132b1..b479bd81120b 100644
> > --- a/drivers/mtd/chips/jedec_probe.c
> > +++ b/drivers/mtd/chips/jedec_probe.c
> > @@ -1889,6 +1889,8 @@ static inline u32 jedec_read_mfr(struct map_info *map,
> > uint32_t base, do {
> >  		uint32_t ofs = cfi_build_cmd_addr(0 + (bank << 8), map, cfi);
> >  		mask = (1 << (cfi->device_type * 8)) - 1;
> > +		if (ofs >= map->size)
> > +			return 0;
> >  		result = map_read(map, base + ofs);
> >  		bank++;
> >  	} while ((result.x[0] & mask) == CFI_MFR_CONTINUATION);  
> 
> The fix is legit but I'm not sure whether we should emit a warning in this 
> case too since something is obviously wrong.
> Boris?

Looks good to me: 0 does not seem to be a valid id, so it should not
allow the caller to find a valid match (we could also return
CFI_MFR_CONTINUATION but 0 is fine).

Linus, do you want to add a Fixes and Cc-stable tag so that it can be
backported to stable kernels?


-- 
Boris Brezillon, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com