[v5.4 stable] arm: stm32: Regression observed on "no-map" reserved memory region

[Florian Fainelli]

On 4/20/2021 9:10 AM, Ard Biesheuvel wrote:
> On Tue, 20 Apr 2021 at 17:54, Rob Herring <robh+dt at kernel.org> wrote:
>>
>> On Tue, Apr 20, 2021 at 10:12 AM Alexandre TORGUE
>> <alexandre.torgue at foss.st.com> wrote:
>>>
>>>
>>>
>>> On 4/20/21 4:45 PM, Rob Herring wrote:
>>>> On Tue, Apr 20, 2021 at 9:03 AM Alexandre TORGUE
>>>> <alexandre.torgue at foss.st.com> wrote:
>>>>>
>>>>> Hi,
>>>>
>>>> Greg or Sasha won't know what to do with this. Not sure who follows
>>>> the stable list either. Quentin sent the patch, but is not the author.
>>>> Given the patch in question is about consistency between EFI memory
>>>> map boot and DT memory map boot, copying EFI knowledgeable folks would
>>>> help (Ard B for starters).
>>>
>>> Ok thanks for the tips. I add Ard in the loop.
>>
>> Sigh. If it was only Ard I was suggesting I would have done that
>> myself. Now everyone on the patch in question and relevant lists are
>> Cc'ed.
>>
> 
> Thanks for the cc.
> 
>>>
>>> Ard, let me know if other people have to be directly added or if I have
>>> to resend to another mailing list.
>>>
>>> thanks
>>> alex
>>>
>>>>
>>>>>
>>>>> Since v5.4.102 I observe a regression on stm32mp1 platform: "no-map"
>>>>> reserved-memory regions are no more "reserved" and make part of the
>>>>> kernel System RAM. This causes allocation failure for devices which try
>>>>> to take a reserved-memory region.
>>>>>
>>>>> It has been introduced by the following path:
>>>>>
>>>>> "fdt: Properly handle "no-map" field in the memory region
>>>>> [ Upstream commit 86588296acbfb1591e92ba60221e95677ecadb43 ]"
>>>>> which replace memblock_remove by memblock_mark_nomap in no-map case.
>>>>>
> 
> Why was this backported? It doesn't look like a bugfix to me.
> 
>>>>> Reverting this patch it's fine.
>>>>>
>>>>> I add part of my DT (something is maybe wrong inside):
>>>>>
>>>>> memory at c0000000 {
>>>>>          reg = <0xc0000000 0x20000000>;
>>>>> };
>>>>>
>>>>> reserved-memory {
>>>>>          #address-cells = <1>;
>>>>>          #size-cells = <1>;
>>>>>          ranges;
>>>>>
>>>>>          gpu_reserved: gpu at d4000000 {
>>>>>                  reg = <0xd4000000 0x4000000>;
>>>>>                  no-map;
>>>>>          };
>>>>> };
>>>>>
>>>>> Sorry if this issue has already been raised and discussed.
>>>>>
> 
> Could you explain why it fails? The region is clearly part of system
> memory, and tagged as no-map, so the patch in itself is not
> unreasonable. However, we obviously have code that relies on how the
> region is represented in /proc/iomem, so it would be helpful to get
> some insight into why this is the case.

I do wonder as well, we have a 32MB "no-map" reserved memory region on
our platforms located at 0xfe000000. Without the offending commit,
/proc/iomem looks like this:

40000000-fdffefff : System RAM
  40008000-40ffffff : Kernel code
  41e00000-41ef1d77 : Kernel data
100000000-13fffffff : System RAM

and with the patch applied, we have this:

40000000-fdffefff : System RAM
  40008000-40ffffff : Kernel code
  41e00000-41ef3db7 : Kernel data
fdfff000-ffffffff : System RAM
100000000-13fffffff : System RAM

so we can now see that the region 0xfe000000 - 0xfffffff is also cobbled
up with the preceding region which is a mailbox between Linux and the
secure monitor at 0xfdfff000 and of size 4KB. It seems like there is

The memblock=debug outputs is also different:

[    0.000000] MEMBLOCK configuration:
[    0.000000]  memory size = 0xfdfff000 reserved size = 0x7ce4d20d
[    0.000000]  memory.cnt  = 0x2
[    0.000000]  memory[0x0]     [0x00000040000000-0x000000fdffefff],
0xbdfff000 bytes flags: 0x0
[    0.000000]  memory[0x1]     [0x00000100000000-0x0000013fffffff],
0x40000000 bytes flags: 0x0
[    0.000000]  reserved.cnt  = 0x6
[    0.000000]  reserved[0x0]   [0x00000040003000-0x0000004000e494],
0xb495 bytes flags: 0x0
[    0.000000]  reserved[0x1]   [0x00000040200000-0x00000041ef1d77],
0x1cf1d78 bytes flags: 0x0
[    0.000000]  reserved[0x2]   [0x00000045000000-0x000000450fffff],
0x100000 bytes flags: 0x0
[    0.000000]  reserved[0x3]   [0x00000047000000-0x0000004704ffff],
0x50000 bytes flags: 0x0
[    0.000000]  reserved[0x4]   [0x000000c2c00000-0x000000fdbfffff],
0x3b000000 bytes flags: 0x0
[    0.000000]  reserved[0x5]   [0x00000100000000-0x0000013fffffff],
0x40000000 bytes flags: 0x0

[    0.000000] MEMBLOCK configuration:
[    0.000000]  memory size = 0x100000000 reserved size = 0x7ca4f24d
[    0.000000]  memory.cnt  = 0x3
[    0.000000]  memory[0x0]     [0x00000040000000-0x000000fdffefff],
0xbdfff000 bytes flags: 0x0
[    0.000000]  memory[0x1]     [0x000000fdfff000-0x000000ffffffff],
0x2001000 bytes flags: 0x4
[    0.000000]  memory[0x2]     [0x00000100000000-0x0000013fffffff],
0x40000000 bytes flags: 0x0
[    0.000000]  reserved.cnt  = 0x6
[    0.000000]  reserved[0x0]   [0x00000040003000-0x0000004000e494],
0xb495 bytes flags: 0x0
[    0.000000]  reserved[0x1]   [0x00000040200000-0x00000041ef3db7],
0x1cf3db8 bytes flags: 0x0
[    0.000000]  reserved[0x2]   [0x00000045000000-0x000000450fffff],
0x100000 bytes flags: 0x0
[    0.000000]  reserved[0x3]   [0x00000047000000-0x0000004704ffff],
0x50000 bytes flags: 0x0
[    0.000000]  reserved[0x4]   [0x000000c3000000-0x000000fdbfffff],
0x3ac00000 bytes flags: 0x0
[    0.000000]  reserved[0x5]   [0x00000100000000-0x0000013fffffff],
0x40000000 bytes flags: 0x0

in the second case we can clearly see that the 32MB no-map region is now
considered as usable RAM.

Hope this helps.

> 
> In any case, the mere fact that this causes a regression should be
> sufficient justification to revert/withdraw it from v5.4, as I don't
> see a reason why it was merged there in the first place. (It has no
> fixes tag or cc:stable)

Agreed, however that means we still need to find out whether a more
recent kernel is also broken, I should be able to tell you that a little
later.
-- 
Florian