2022-12-21 06:38:14

by Srinivasa Rao Mandadapu

Subject: [PATCH v4 1/2] remoteproc: elf_loader: Update resource table name check

Update the way of checking resource table name with prefix substring search
instead of complete string search.

In general, Qualcomm DSP binary is prepared by combining different ELFs'.
Hence, section header name (e.g. .resource_table), appended with ELF name
to differentiate with same section(e.g. resource_table.ac_bin_process) of
different ELFs'.

Example readelf output of DSP binary:
[60] .start.ac_bin_process PROGBITS
[61] .resource_table.ac_bin_process PROGBITS
[62] .comment.ac_bin_process PROGBITS

Signed-off-by: Srinivasa Rao Mandadapu <[email protected]>
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Reviewed-by: Mukesh Ojha <[email protected]>
---
drivers/remoteproc/remoteproc_elf_loader.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/remoteproc/remoteproc_elf_loader.c b/drivers/remoteproc/remoteproc_elf_loader.c
index 5a412d7..77330d6 100644
--- a/drivers/remoteproc/remoteproc_elf_loader.c
+++ b/drivers/remoteproc/remoteproc_elf_loader.c
@@ -272,7 +272,7 @@ find_table(struct device *dev, const struct firmware *fw)
u64 offset = elf_shdr_get_sh_offset(class, shdr);
u32 name = elf_shdr_get_sh_name(class, shdr);

- if (strcmp(name_table + name, ".resource_table"))
+ if (!strstarts(name_table + name, ".resource_table"))
continue;

table = (struct resource_table *)(elf_data + offset);
--
2.7.4

2023-01-05 00:04:15

by Mathieu Poirier

Subject: Re: [PATCH v4 1/2] remoteproc: elf_loader: Update resource table name check

On Wed, Dec 21, 2022 at 11:40:32AM +0530, Srinivasa Rao Mandadapu wrote:
> Update the way of checking resource table name with prefix substring search
> instead of complete string search.
>
> In general, Qualcomm DSP binary is prepared by combining different ELFs'.
> Hence, section header name (e.g. .resource_table), appended with ELF name
> to differentiate with same section(e.g. resource_table.ac_bin_process) of
> different ELFs'.
>

Well, they are different ELF files so it is implied the resource section may be
different... And from what I see this is purely a human thing and not
specifically needed by machines.

I can't imagine this is a new situation and as such must ask how it was working
before?

> Example readelf output of DSP binary:
> [60] .start.ac_bin_process PROGBITS
> [61] .resource_table.ac_bin_process PROGBITS

The real question is... Why is this labeled ".resource_table.ac_bin_process"
rather than simply ".resource_table"?

> [62] .comment.ac_bin_process PROGBITS
>
> Signed-off-by: Srinivasa Rao Mandadapu <[email protected]>
> Reviewed-by: Philippe Mathieu-Daud? <[email protected]>
> Reviewed-by: Mukesh Ojha <[email protected]>
> ---
> drivers/remoteproc/remoteproc_elf_loader.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/remoteproc/remoteproc_elf_loader.c b/drivers/remoteproc/remoteproc_elf_loader.c
> index 5a412d7..77330d6 100644
> --- a/drivers/remoteproc/remoteproc_elf_loader.c
> +++ b/drivers/remoteproc/remoteproc_elf_loader.c
> @@ -272,7 +272,7 @@ find_table(struct device *dev, const struct firmware *fw)
> u64 offset = elf_shdr_get_sh_offset(class, shdr);
> u32 name = elf_shdr_get_sh_name(class, shdr);
>
> - if (strcmp(name_table + name, ".resource_table"))
> + if (!strstarts(name_table + name, ".resource_table"))
> continue;
>
> table = (struct resource_table *)(elf_data + offset);
> --
> 2.7.4
>