Re: [PATCH V2] remoteproc: core: Clear table_sz when rproc_shutdown

[Peng Fan <peng.fan@oss.nxp.com>]

On Tue, Apr 08, 2025 at 10:59:58AM -0600, Mathieu Poirier wrote:
>On Tue, 8 Apr 2025 at 09:02, Peng Fan <peng.fan@oss.nxp.com> wrote:
>>
>> On Thu, Apr 03, 2025 at 10:32:39PM +0800, Peng Fan wrote:
>> >Hi Bjorn,
>> >
>> >
>> >Thanks for replying this thread.
>> >
>> >On Wed, Apr 02, 2025 at 08:48:58AM -0500, Bjorn Andersson wrote:
>> >>On Wed, Apr 02, 2025 at 09:43:55AM +0800, Peng Fan wrote:
>> >>> On Tue, Apr 01, 2025 at 10:05:03AM -0600, Mathieu Poirier wrote:
>> >>> >On Tue, Apr 01, 2025 at 09:41:24AM +0800, Peng Fan wrote:
>> >>...
>> >>> >
>> >>> >The core is already checking if @loaded_table is valid in rproc_start(), why
>> >>> >can't that be used instead of adding yet another check?
>> >>>
>> >>> Ah. I was thinking clear table_sz in rpoc_shutdown is an easy approach and
>> >>> could benifit others in case other platforms meet similar issue in future.
>> >>>
>> >>
>> >>I like the general idea of keeping things clean and avoid leaving stale
>> >>data behind.
>> >>
>> >>But clearing table_sz during stop in order to hide the fact that the
>> >>future table_ptr will contain valid data that shouldn't be used, that's
>> >>just a bug waiting to show up again in the future.
>> >
>> >Agree.
>> >
>> >Do you need me to post a fix for
>> >commit efdde3d73ab25ce("remoteproc: core: Clear table_sz when rproc_shutdown")
>> >by clearing table_sz in rproc_fw_boot and rproc_detach as did in this v2?
>> >
>> >To i.MX, the above in-tree patch is ok, so all it fine, and this v2 patch
>> >could be dropped.
>> >
>> >But anyway, if you prefer a follow up fix, please let me know, I
>> >could post a patch.
>>
>> Hi Bjorn, Mathieu,
>>
>>  I will wait for one more week to see if any concerns or questions.
>>  Please raise if you have.
>>
>
>I am working with Bjorn to get your patch reverted.  Once that has
>happened you can send another patch.

ok, I am fine with this.

when get reverted, I need use another method to fix the issue.

I posted two approaches[1], but not get you reply. Since Bjorn raised
his concern on 1st approach, I think I need to use the 2nd approach without
touching the core code.
pasted here,
"The 2nd approach is to clear rproc->table_sz and rproc->table_ptr in
imx_rproc_parse_fw before rproc_elf_load_rsc_table.
"

Or a V3 of current patch with updated commit log.

Please suggest.

If you still have concern or things still not clear to you, please
let me know.

[1] https://lore.kernel.org/all/20250402014355.GA22575@nxa18884-linux/

Regards,
Peng

>
>>  If no, I suppose this thread is done and I will start my other work
>>  regarding rproc.
>>
>> Thanks,
>> Peng
>>
>> >
>> >Thanks,
>> >Peng
>> >
>> >>
>> >>Regards,
>> >>Bjorn
>> >>
>> >
>