Re: [Minios-devel] [UNIKRAFT PATCHv4 6/9] plat/common: Introduce fdt_get_address helper

[Sharan Santhanam <sharan.santhanam@xxxxxxxxx>]

Hello Jia He,

Please find the comments inline.

Thanks & Regards
Sharan

On 3/27/19 3:34 AM, Jia He wrote:
> From: Wei Chen <wei.chen@xxxxxxx>
> This helper will be used very frequently for device libraries
> to parse their addresses. Introduce this helper to avoid using
> fdt_address_cells and fdt_size_cells everywhere.
>
> Signed-off-by: Wei Chen <wei.chen@xxxxxxx>
> Signed-off-by: Jianyong Wu <jianyong.wu@xxxxxxx>
> Signed-off-by: Jia He <justin.he@xxxxxxx>
> ---
>   plat/common/fdt.c         | 35 +++++++++++++++++++++++++++++++++++
>   plat/common/include/fdt.h | 18 ++++++++++++++++++
>   2 files changed, 53 insertions(+)
>
> diff --git a/plat/common/fdt.c b/plat/common/fdt.c
> index 1246dec..6cbedec 100644
> --- a/plat/common/fdt.c
> +++ b/plat/common/fdt.c
> @@ -292,3 +292,38 @@ bail:
>         return result;
>   }
>   
> +int fdt_get_address(const void *fdt, int nodeoffset, uint32_t index,
> +                       uint64_t *addr, uint64_t *size)
> +{
> +       int len, prop_addr, prop_size;
> +       int naddr, nsize, term_size;
> +       const void *regs;
UK_ASSERT(addr && size);

> +
> +       naddr = fdt_address_cells_or_parent(fdt, nodeoffset);
Shouldn't this refer to the nodeoffset be the offset of the parent?
In our current implementation of the fdt_address_cells_or_parent we
search for the property of the #address-cell in the node in "nodeoffset" 
which should offset of its parent.
> +       if (naddr < 0 || naddr >= FDT_MAX_NCELLS)
> +               return -FDT_ERR_BADNCELLS;
> +
Shouldn't this refer to the nodeoffset be the offset of the parent?
In our current implementation of the fdt_address_cells_or_parent we
search for the property of the #size-cell in the node in "nodeoffset" 
which should offset of its parent.
> +       nsize = fdt_size_cells_or_parent(fdt, nodeoffset);
> +       if (nsize < 0 || nsize >= FDT_MAX_NCELLS)
> +               return -FDT_ERR_BADNCELLS;
> +
> +       /* Get reg content */
> +       regs = fdt_getprop(fdt, nodeoffset, "reg", &len);
Instead of returning -FDT_ERR_NOTFOUND, we could return len as it 
contains the error code.
> +       if (regs == NULL)
> +               return -FDT_ERR_NOTFOUND;
> +
> +       term_size = sizeof(fdt32_t) * (nsize + naddr);
> +       prop_addr = term_size * index;
> +       prop_size = prop_addr + sizeof(fdt32_t) * naddr;
> +
> +       /* The reg content must cover the reg term[index] at least */
> +       if (len < (prop_addr + term_size))
> +               return -FDT_ERR_NOSPACE;
> +
> +       *size = fdt_reg_read_number(regs + prop_size, nsize);

We return FDT_BAD_ADDR but the return code is 0. Is this behavior expected?
> +       /* Handle ranges property, currently only 1:1 mapping is supported */
> +       *addr = fdt_translate_address_by_ranges(fdt, nodeoffset,
> +                                               regs + prop_addr);
> +
> +       return 0;
> +}
> diff --git a/plat/common/include/fdt.h b/plat/common/include/fdt.h
> index 969cd28..27f4f2c 100644
> --- a/plat/common/include/fdt.h
> +++ b/plat/common/include/fdt.h
> @@ -110,4 +110,22 @@ int fdt_address_cells_or_parent(const void *fdt, int 
> nodeoffset);
>    */
>   int fdt_size_cells_or_parent(const void *fdt, int nodeoffset);
>   
> +/**
> + * fdt_get_address - retrieve device address of a given index
> + * @fdt: pointer to the device tree blob
> + * @nodeoffset: offset of the node to find the address for.
> + * @index: index of region
> + * @addr: return the region address
> + * @size: return the region size
> + *
> + * returns:
> + *     0, on success
> + *      -FDT_ERR_BADNCELLS, if the node has a badly formatted or invalid
> + *             address property
> + *      -FDT_ERR_NOTFOUND, if the node doesn't have address property
> + *      -FDT_ERR_NOSPACE, if the node doesn't have address for index
> + */
> +int fdt_get_address(const void *fdt, int nodeoffset, int index,
> +                       uint64_t *addr, uint64_t *size);
> +
>   #endif
_______________________________________________
Minios-devel mailing list
Minios-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/minios-devel