Re: [PATCH v10 03/12] cxl: Share HDM decoder decode logic

[Srirangan Madhavan <smadhavan@nvidia.com>]

On 8/20/26 4:25 PM, Dave Jiang wrote:
> I'm concerned about here modifying cxled->skip without the DPA write lock. Should it be:
>          skip = settings.target_or_skip;
> 
> And leave modification of cxled->skip to exclusively with devm_cxl_dpa_reserve()?
> 
> Also this change isn't mentioned by the commit log.
> 
> DJ
Thank you, addressed in v11.

   - Changed the decoder-count check to return bool and renamed the 
internal and wrapper helpers to __cxl_pci_hdm_decoder_count_match() and 
cxl_pci_hdm_decoder_count_match().
   - Renamed the switch helper to cxl_switch_get_target_list().
   - Renamed the cache update helper to cxl_hdm_save_decoder_info().
   - Initialize the settings ID before the enabled-decoder check, 
removing the multi-line else.
   - Added a lock assertion to cxl_decoder_snapshot() using a 
lockdep_assert_held().

   After the refactor, the snapshot and switch-target changes now appear 
in patch 1. The PCI HDM cache updates remain in patch 4.

-- 
Regards,
Srirangan