lkml.org 
[lkml]   [2022]   [Oct]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v4 3/4] fpga: dfl: add basic support DFHv1


On Fri, 21 Oct 2022, Andy Shevchenko wrote:

> On Thu, Oct 20, 2022 at 02:26:09PM -0700, matthew.gerlach@linux.intel.com wrote:
>> From: Matthew Gerlach <matthew.gerlach@linux.intel.com>
>>
>> Add generic support for MSI-X interrupts for DFL devices.
>>
>> The location of a feature's registers is explicitly
>> described in DFHv1 and can be relative to the base of the DFHv1
>> or an absolute address. Parse the location and pass the information
>> to DFL driver.
>
> ...
>
>> +static void *find_param(void *base, resource_size_t max, int param)
>
> Why base can't be u64 * to begin with?

It can be u64, and I will consider it for the next iteration.
>
>> +{
>> + int off = 0;
>> + u64 v, next;
>> +
>> + while (off < max) {
>
> Maybe you need a comment somewhere to tell that the caller guarantees that max
> won't provoke OOB accesses.
>
>> + v = *(u64 *)(base + off);
>
> Okay, if offset is not multiple of at least 4, how do you guarantee no
> exception on the architectures with disallowed misaligned accesses?
>
> Making base to be u64 * solves this, but you need to take care to provide
> offset in terms of u64 words.

The masking of next below ensures that the offset it at least 4 byte
aligned, but it might make sense to define the next field in terms of 8
byte words.

>
>> + if (param == FIELD_GET(DFHv1_PARAM_HDR_ID, v))
>> + return base + off + DFHv1_PARAM_DATA;
>> +
>> + next = FIELD_GET(DFHv1_PARAM_HDR_NEXT_OFFSET, v);
>> + off += next & ~DFHv1_PARAM_HDR_NEXT_MASK;
>> + if (next & DFHv1_PARAM_HDR_NEXT_EOL)
>> + break;
>> +
>> + }
>> +
>> + return NULL;
>> +}
>
> ...
>
>> + /*
>> + * DFHv0 only provides mmio resource information for each feature
>
> MMIO

I'll change mmio to MMIO here and a place in the documentation that I
noticed.

>
>> + * in the DFL header. There is no generic interrupt information.
>> + * Instead, features with interrupt functionality provide
>> + * the information in feature specific registers.
>> + */
>
> ...
>
>> + if (!finfo->param_size)
>> break;
>
> This is redundant as it's implied by find_param().

I will remove the redundant code.

>
>> + p = find_param(params, finfo->param_size, DFHv1_PARAM_ID_MSI_X);
>> + if (!p)
>> break;
>
> ...
>
>> +static int dfh_get_psize(void __iomem *dfh_base, resource_size_t max)
>> +{
>> + int size = 0;
>> + u64 v, next;
>> +
>> + if (!FIELD_GET(DFHv1_CSR_SIZE_GRP_HAS_PARAMS,
>> + readq(dfh_base + DFHv1_CSR_SIZE_GRP)))
>> + return 0;
>> +
>> + while (size + DFHv1_PARAM_HDR < max) {
>> + v = readq(dfh_base + DFHv1_PARAM_HDR + size);
>> +
>> + next = FIELD_GET(DFHv1_PARAM_HDR_NEXT_OFFSET, v);
>> + if (!(next & ~DFHv1_PARAM_HDR_NEXT_MASK))
>> + return -EINVAL;
>> +
>> + size += next & ~DFHv1_PARAM_HDR_NEXT_MASK;
>> +
>> + if (next & DFHv1_PARAM_HDR_NEXT_EOL)
>> + return size;
>
> These 3 looks like they deserve different fields and hence separate FIELD_GET()
> will return exactly what we need without additional masking, right?

I agree separate FIELD_GET() calls will be cleaner.

>
>> + }
>> +
>> + return -ENOENT;
>> +}
>
> ...
>
>> + if (dfh_psize > 0) {
>
> Isn't this implied by memcpy_fromio()? I mean if it's 0, nothing bad will
> happen if you call the above directly.
>
>> + memcpy_fromio(finfo->params,
>> + binfo->ioaddr + ofst + DFHv1_PARAM_HDR, dfh_psize);
>> + finfo->param_size = dfh_psize;
>> + }
>
> ...
>
>> finfo->mmio_res.flags = IORESOURCE_MEM;
>> + if (dfh_ver == 1) {
>> + v = readq(binfo->ioaddr + ofst + DFHv1_CSR_ADDR);
>> + if (v & DFHv1_CSR_ADDR_REL)
>> + finfo->mmio_res.start = v & ~DFHv1_CSR_ADDR_REL;
>> + else
>> + finfo->mmio_res.start = binfo->start + ofst +
>> + FIELD_GET(DFHv1_CSR_ADDR_MASK, v);
>> +
>> + v = readq(binfo->ioaddr + ofst + DFHv1_CSR_SIZE_GRP);
>> + finfo->mmio_res.end = finfo->mmio_res.start +
>> + FIELD_GET(DFHv1_CSR_SIZE_GRP_SIZE, v) - 1;
>> + } else {
>> + finfo->mmio_res.start = binfo->start + ofst;
>> + finfo->mmio_res.end = finfo->mmio_res.start + size - 1;
>> + }
>
> You may define
>
> resource_size_t start, end;
>
> locally and simplify above quite a bit.

That is a good suggestion that should clean up the code quite a bit.

>
> ...
>
>> +void *dfh_find_param(struct dfl_device *dfl_dev, int param);
>
> + Blank line.
>
>> #endif /* __LINUX_DFL_H */
>
> --
> With Best Regards,
> Andy Shevchenko
>
>
>

\
 
 \ /
  Last update: 2022-10-24 18:26    [W:0.114 / U:1.112 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site