Message-ID: <9959cb08-5137-709b-faf6-46fee024630d@arm.com>
Date:   Tue, 18 Oct 2022 14:06:05 +0100
From:   Robin Murphy <robin.murphy@....com>
To:     Jianmin Lv <lvjianmin@...ngson.cn>,
        "Rafael J. Wysocki" <rafael@...nel.org>
Cc:     Yicong Yang <yangyicong@...wei.com>, yangyicong@...ilicon.com,
        lpieralisi@...nel.org, chenhuacai@...ngson.cn,
        guohanjun@...wei.com, sudeep.holla@....com, lenb@...nel.org,
        robert.moore@...el.com, linux-kernel@...r.kernel.org,
        linux-acpi@...r.kernel.org, loongarch@...ts.linux.dev,
        liulongfang <liulongfang@...wei.com>
Subject: Re: [PATCH V5 1/2] ACPI / scan: Support multiple dma windows with
 different offsets

On 2022-10-18 14:00, Jianmin Lv wrote:
> 
> 
> On 2022/10/18 下午8:56, Rafael J. Wysocki wrote:
>> On Tue, Oct 18, 2022 at 2:50 PM Jianmin Lv <lvjianmin@...ngson.cn> wrote:
>>>
>>>
>>>
>>> On 2022/10/18 下午8:32, Yicong Yang wrote:
>>>> On 2022/10/18 20:00, Robin Murphy wrote:
>>>>> On 2022-10-18 11:08, Rafael J. Wysocki wrote:
>>>>>> On Tue, Oct 18, 2022 at 11:33 AM Yicong Yang 
>>>>>> <yangyicong@...wei.com> wrote:
>>>>>>>
>>>>>>> On 2022/9/11 17:06, Jianmin Lv wrote:
>>>>>>>> In DT systems configurations, of_dma_get_range() returns struct
>>>>>>>> bus_dma_region DMA regions; they are used to set-up devices
>>>>>>>> DMA windows with different offset available for translation 
>>>>>>>> between DMA
>>>>>>>> address and CPU address.
>>>>>>>>
>>>>>>>> In ACPI systems configuration, acpi_dma_get_range() does not return
>>>>>>>> DMA regions yet and that precludes setting up the 
>>>>>>>> dev->dma_range_map
>>>>>>>> pointer and therefore DMA regions with multiple offsets.
>>>>>>>>
>>>>>>>> Update acpi_dma_get_range() to return struct bus_dma_region
>>>>>>>> DMA regions like of_dma_get_range() does.
>>>>>>>>
>>>>>>>> After updating acpi_dma_get_range(), acpi_arch_dma_setup() is 
>>>>>>>> changed for
>>>>>>>> ARM64, where the original dma_addr and size are removed as these
>>>>>>>> arguments are now redundant, and pass 0 and U64_MAX for dma_base
>>>>>>>> and size of arch_setup_dma_ops; this is a simplification consistent
>>>>>>>> with what other ACPI architectures also pass to 
>>>>>>>> iommu_setup_dma_ops().
>>>>>>>>
>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> With this patch we met problem as well. The DMA coherent mask is 
>>>>>>> not set correctly
>>>>>>> for a ehci usb controller and lead to the below calltrace:
>>>>>>>
>>>>>>> [   16.699259] ------------[ cut here ]------------
>>>>>>> [   16.703855] WARNING: CPU: 0 PID: 853 at 
>>>>>>> kernel/dma/mapping.c:499 dma_alloc_attrs+0xc0/0xf0
>>>>>>> [   16.712082] Modules linked in:
>>>>>>> [   16.715124] CPU: 0 PID: 853 Comm: kworker/0:3 Not tainted 
>>>>>>> 6.1.0-rc1-pipe-deadlock+ #5
>>>>>>> [   16.722916] Hardware name: Huawei TaiShan 2280 V2/BC82AMDC, 
>>>>>>> BIOS 2280-V2 CS V5.B211.01 11/10/2021
>>>>>>> [   16.731745] Workqueue: events work_for_cpu_fn
>>>>>>> [   16.736083] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT 
>>>>>>> -SSBS BTYPE=--)
>>>>>>> [   16.743013] pc : dma_alloc_attrs+0xc0/0xf0
>>>>>>> [   16.747091] lr : dma_pool_alloc+0x11c/0x200
>>>>>>> [   16.751255] sp : ffff80001e46bb50
>>>>>>> [   16.754554] x29: ffff80001e46bb50 x28: 0000000000000000 x27: 
>>>>>>> 0000000000000000
>>>>>>> [   16.761657] x26: ffff80000b33ce18 x25: ffff800009cc6c48 x24: 
>>>>>>> 0000000000000000
>>>>>>> [   16.768759] x23: ffff00208c830918 x22: 0000000000001000 x21: 
>>>>>>> 0000000000000cc0
>>>>>>> [   16.775861] x20: ffff00208ae82080 x19: ffff0020865c40d0 x18: 
>>>>>>> 0000000000000030
>>>>>>> [   16.782964] x17: 626d756e20737562 x16: 2064656e67697373 x15: 
>>>>>>> ffff00208ae82640
>>>>>>> [   16.790066] x14: 0000000000000000 x13: 646e756f72616b72 x12: 
>>>>>>> 6f77204348207379
>>>>>>> [   16.797167] x11: 73706f6e79532067 x10: ffff205f43980000 x9 : 
>>>>>>> ffff80000830b3ac
>>>>>>> [   16.804269] x8 : ffff0020861b1b00 x7 : 0000000000000000 x6 : 
>>>>>>> 0000000000000000
>>>>>>> [   16.811371] x5 : 0000000000000000 x4 : 0000000000000000 x3 : 
>>>>>>> 0000000000000cc0
>>>>>>> [   16.818472] x2 : ffff00208c830918 x1 : 0000000000001000 x0 : 
>>>>>>> 0000000000000000
>>>>>>> [   16.825574] Call trace:
>>>>>>> [   16.828009]  dma_alloc_attrs+0xc0/0xf0
>>>>>>> [   16.831741]  dma_pool_alloc+0x11c/0x200
>>>>>>> [   16.835559]  ehci_qh_alloc+0x60/0x12c
>>>>>>> [   16.839207]  ehci_setup+0x18c/0x40c
>>>>>>> [   16.842680]  ehci_pci_setup+0xb8/0x680
>>>>>>> [   16.846412]  usb_add_hcd+0x310/0x5c0
>>>>>>> [   16.849973]  usb_hcd_pci_probe+0x254/0x36c
>>>>>>> [   16.854051]  ehci_pci_probe+0x40/0x60
>>>>>>> [   16.857698]  local_pci_probe+0x48/0xb4
>>>>>>> [   16.861431]  work_for_cpu_fn+0x24/0x40
>>>>>>> [   16.865163]  process_one_work+0x1e0/0x450
>>>>>>> [   16.869155]  worker_thread+0x2cc/0x44c
>>>>>>> [   16.872886]  kthread+0x114/0x120
>>>>>>> [   16.876099]  ret_from_fork+0x10/0x20
>>>>>>> [   16.879657] ---[ end trace 0000000000000000 ]---
>>>>>>>
>>>>>>> After reverting this patch the problem resolved. Tested on the 
>>>>>>> latest 6.1-rc1.
>>>>>>
>>>>>> OK, I'll queue up a revert of this and one more commit depending 
>>>>>> on it.
>>>>>
>>>>> FWIW it looks like the fix should be as simple as below.
>>>>>
>>>>
>>>> Looks like it's the case. The change works on my platform, now the 
>>>> ehci probed successfully again
>>>> with no calltrace:
>>>>
>>>> Tested-by: Yicong Yang <yangyicong@...ilicon.com>
>>>>
>>>>> Robin.
>>>>>
>>>>> ----->8-----
>>>>> diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
>>>>> index 558664d169fc..b6962bff1eae 100644
>>>>> --- a/drivers/acpi/scan.c
>>>>> +++ b/drivers/acpi/scan.c
>>>>> @@ -1509,6 +1509,7 @@ int acpi_dma_get_range(struct device *dev, 
>>>>> const struct bus_dma_region **map)
>>>>>                goto out;
>>>>>            }
>>>>>
>>>>> +        *map = r;
>>>>>            list_for_each_entry(rentry, &list, node) {
>>>>>                if (rentry->res->start >= rentry->res->end) {
>>>>>                    kfree(r);
>>>>> @@ -1523,8 +1524,6 @@ int acpi_dma_get_range(struct device *dev, 
>>>>> const struct bus_dma_region **map)
>>>>>                r->offset = rentry->offset;
>>>>>                r++;
>>>>>            }
>>>>> -
>>>>> -        *map = r;
>>>>>        }
>>>>>     out:
>>>>>        acpi_dev_free_resource_list(&list);
>>>>>
>>>
>>> Ohh, yes, map got a wrong value of r because it has been changed.
>>
>> Well, please send me a working patch by EOD tomorrow.
>>
> 
> Ok!

In fact there's another related bug in the error path as well. I'm 
writing up the patch now...

Thanks,
Robin

> 
> 
>>> Maybe wo can fix it like this:
>>>
>>> truct bus_dma_region *r, *orig_r;
>>> ...
>>> orig_r = r = kcalloc(ret + 1, sizeof(*r), GFP_KERNEL);
>>> ...
>>> *map = orig_r;
>>>
>>>>> .
> 

Powered by blists - more mailing lists