mtd: jedec_probe: Fix crash in jedec_read_mfr()

Message ID 20180303222903.27767-1-linus.walleij@linaro.org
State Accepted
Delegated to: Boris Brezillon
Headers show
Series
  • mtd: jedec_probe: Fix crash in jedec_read_mfr()
Related show

Commit Message

Linus Walleij March 3, 2018, 10:29 p.m.
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@linaro.org>
---
 drivers/mtd/chips/jedec_probe.c | 2 ++
 1 file changed, 2 insertions(+)

Comments

Linus Walleij March 27, 2018, 5:51 p.m. | #1
On Sat, Mar 3, 2018 at 11:29 PM, Linus Walleij <linus.walleij@linaro.org> wrote:

> 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@linaro.org>

Could someone maintaining MTD please pick up this bug fix?

It is a very real problem to me.

Yours,
Linus Walleij
Richard Weinberger March 27, 2018, 7:02 p.m. | #2
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@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?

Thanks,
//richard
Linus Walleij March 27, 2018, 8:31 p.m. | #3
On Tue, Mar 27, 2018 at 9:02 PM, Richard Weinberger <richard@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:
>>
>>               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.

What is wrong is perhaps our way of probing JEDEC
Flashes: the condition is hit when accessing a device on a
16bit bus in 8bit mode, which is what the probe code tries
before it tries 16bit.

However there are clear comments in the
code (above jedec_match()) indicating that it is impossible
to do a perfect probe routine so this could be expected to
happen to some extent,I am merely amending an issue that
may be unfixable.

Yours,
Linus Walleij
Boris Brezillon March 28, 2018, 7:03 a.m. | #4
On Tue, 27 Mar 2018 21:02:08 +0200
Richard Weinberger <richard@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@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 March 29, 2018, 5:24 p.m. | #5
On Sat,  3 Mar 2018 23:29:03 +0100
Linus Walleij <linus.walleij@linaro.org> wrote:

> 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@linaro.org>

Applied.

Also added:

Fixes: 5c9c11e1c47c ("[MTD] [NOR] Add support for flash chips with ID in bank other than 0")
Cc: <stable@vger.kernel.org>

I'll send a fixes PR to Linus soon.

Thanks,

Boris

> ---
>  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);

Patch

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);