Re: [BUG 5.14] arm64/mm: dma memory mapping fails (in some cases)

From: David Hildenbrand <david@redhat.com>
To: Robin Murphy <robin.murphy@arm.com>,
	Mike Rapoport <rppt@kernel.org>,
	Catalin Marinas <catalin.marinas@arm.com>
Cc: Alex Bee <knaerzche@gmail.com>, Will Deacon <will@kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Anshuman Khandual <anshuman.khandual@arm.com>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	linux-mm@kvack.org,
	Linux ARM <linux-arm-kernel@lists.infradead.org>
Subject: Re: [BUG 5.14] arm64/mm: dma memory mapping fails (in some cases)
Date: Tue, 24 Aug 2021 20:59:22 +0200	[thread overview]
Message-ID: <60a11eba-2910-3b5f-ef96-97d4556c1596@redhat.com> (raw)
In-Reply-To: <0908ce39-7e30-91fa-68ef-11620f9596ae@arm.com>

On 24.08.21 20:46, Robin Murphy wrote:
> On 2021-08-24 19:28, Mike Rapoport wrote:
>> On Tue, Aug 24, 2021 at 06:37:41PM +0100, Catalin Marinas wrote:
>>> Hi Alex,
>>>
>>> Thanks for the report.
>>>
>>> On Tue, Aug 24, 2021 at 03:40:47PM +0200, Alex Bee wrote:
>>>> it seems there is a regression in arm64 memory mapping in 5.14, since it
>>>> fails on Rockchip RK3328 when the pl330 dmac tries to map with:
>>>>
>>>> [ï¿½ï¿½ï¿½ 8.921909] ------------[ cut here ]------------
>>>> [ï¿½ï¿½ï¿½ 8.921940] WARNING: CPU: 2 PID: 373 at kernel/dma/mapping.c:235 dma_map_resource+0x68/0xc0
>>>> [ï¿½ï¿½ï¿½ 8.921973] Modules linked in: spi_rockchip(+) fuse
>>>> [ï¿½ï¿½ï¿½ 8.921996] CPU: 2 PID: 373 Comm: systemd-udevd Not tainted 5.14.0-rc7 #1
>>>> [ï¿½ï¿½ï¿½ 8.922004] Hardware name: Pine64 Rock64 (DT)
>>>> [ï¿½ï¿½ï¿½ 8.922011] pstate: 80000005 (Nzcv daif -PAN -UAO -TCO BTYPE=--)
>>>> [ï¿½ï¿½ï¿½ 8.922018] pc : dma_map_resource+0x68/0xc0
>>>> [ï¿½ï¿½ï¿½ 8.922026] lr : pl330_prep_slave_fifo+0x78/0xd0
>>>> [ï¿½ï¿½ï¿½ 8.922040] sp : ffff800012102ae0
>>>> [ï¿½ï¿½ï¿½ 8.922043] x29: ffff800012102ae0 x28: ffff000005c94800 x27: 0000000000000000
>>>> [ï¿½ï¿½ï¿½ 8.922056] x26: ffff000000566bd0 x25: 0000000000000001 x24: 0000000000000001
>>>> [ï¿½ï¿½ï¿½ 8.922067] x23: 0000000000000002 x22: ffff000000628c00 x21: 0000000000000001
>>>> [ï¿½ï¿½ï¿½ 8.922078] x20: ffff000000566bd0 x19: 0000000000000001 x18: 0000000000000000
>>>> [ï¿½ï¿½ï¿½ 8.922089] x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
>>>> [ï¿½ï¿½ï¿½ 8.922100] x14: 0000000000000277 x13: 0000000000000001 x12: 0000000000000000
>>>> [ï¿½ï¿½ï¿½ 8.922111] x11: 0000000000000001 x10: 00000000000008e0 x9 : ffff800012102a80
>>>> [ï¿½ï¿½ï¿½ 8.922123] x8 : ffff000000d14b80 x7 : ffff0000fe7b12f0 x6 : ffff0000fe7b1100
>>>> [ï¿½ï¿½ï¿½ 8.922134] x5 : fffffc000000000f x4 : 0000000000000000 x3 : 0000000000000001
>>>> [ï¿½ï¿½ï¿½ 8.922145] x2 : 0000000000000001 x1 : 00000000ff190800 x0 : ffff000000628c00
>>>> [ï¿½ï¿½ï¿½ 8.922158] Call trace:
>>>> [ï¿½ï¿½ï¿½ 8.922163]ï¿½ dma_map_resource+0x68/0xc0
>>>> [ï¿½ï¿½ï¿½ 8.922173]ï¿½ pl330_prep_slave_sg+0x58/0x220
>>>> [ï¿½ï¿½ï¿½ 8.922181]ï¿½ rockchip_spi_prepare_dma+0xd8/0x2c0 [spi_rockchip]
>>>> [ï¿½ï¿½ï¿½ 8.922208]ï¿½ rockchip_spi_transfer_one+0x294/0x3d8 [spi_rockchip]
>>> [...]
>>>> Note: This does not relate to the spi driver - when disabling this device in
>>>> the device tree it fails for any other (i2s, for instance) which uses dma.
>>>> Commenting out the failing check at [1], however, helps and the mapping
>>>> works again.
>>
>>> Do you know which address dma_map_resource() is trying to map (maybe
>>> add some printk())? It's not supposed to map RAM, hence the warning.
>>> Random guess, the address is 0xff190800 (based on the x1 above but the
>>> regs might as well be mangled).
>>
>> 0xff190800 will cause this warning for sure. It has a memory map, but it is
>> not RAM so old version of pfn_valid() would return 0 and the new one
>> returns 1.
> 
> How does that happen, though? It's not a memory address, and it's not
> even within the bounds of anywhere there should or could be memory. This
> SoC has a simple memory map - everything from 0 to 0xfeffffff goes to
> the DRAM controller (which may not all be populated, and may have pieces
> carved out by secure firmware), while 0xff000000-0xffffffff is MMIO. Why
> do we have pages (or at least the assumption of pages) for somewhere
> which by all rights should not have them?

Simple: we allocate the vmemmap for whole sections (e.g., 128 MiB) to 
avoid any such hacks. If there is a memory hole, it gets a memmap as well.

Tricking pfn_valid() into returning "false" where we actually have a 
memmap only makes it look like there is no memmap; but there is one, and
it's PG_reserved.

[...]

>>> Either pfn_valid() gets confused in 5.14 or something is wrong with the
>>> DT. I have a suspicion it's the former since reverting the above commit
>>> makes it disappear.
>>
>> I think pfn_valid() actually behaves as expected but the caller is wrong
>> because pfn_valid != RAM (this applies btw to !arm64 as well).
>>
>> 	/* Don't allow RAM to be mapped */
>> 	if (WARN_ON_ONCE(pfn_valid(PHYS_PFN(phys_addr))))
>> 		return DMA_MAPPING_ERROR;
>>
>> Alex, can you please try this patch:
> 
> That will certainly paper over the issue, but it's avoiding the question
> of what went wrong with the memory map in the first place. The comment
> is indeed a bit inaccurate, but ultimately dma_map_resource() exists for
> addresses that would be wrong to pass to dma_map_page(), so I believe
> pfn_valid() is still the correct check.

If we want to check for RAM, pfn_valid() would be wrong. If we want to 
check for "is there a memmap, for whatever lives or does not live 
there", pfn_valid() is the right check.

-- 
Thanks,

David / dhildenb