From mboxrd@z Thu Jan 1 00:00:00 1970 From: jeremy.linton@arm.com (Jeremy Linton) Date: Wed, 17 Jan 2018 12:51:31 -0600 Subject: [PATCH v6 02/12] drivers: base: cacheinfo: setup DT cache properties early In-Reply-To: References: <20180113005920.28658-1-jeremy.linton@arm.com> <20180113005920.28658-3-jeremy.linton@arm.com> <20180115123338.GB5473@e107155-lin> <65f78c99-8b86-0098-7ced-899840a4bf16@arm.com> Message-ID: <15884ccc-2cfb-da91-5844-369d8237175d@arm.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Hi, On 01/17/2018 12:20 PM, Sudeep Holla wrote: > > > On 16/01/18 21:07, Jeremy Linton wrote: >> Hi, >> >> On 01/15/2018 06:33 AM, Sudeep Holla wrote: >>> On Fri, Jan 12, 2018 at 06:59:10PM -0600, Jeremy Linton wrote: >>>> The original intent in cacheinfo was that an architecture >>>> specific populate_cache_leaves() would probe the hardware >>>> and then cache_shared_cpu_map_setup() and >>>> cache_override_properties() would provide firmware help to >>>> extend/expand upon what was probed. Arm64 was really >>>> the only architecture that was working this way, and >>>> with the removal of most of the hardware probing logic it >>>> became clear that it was possible to simplify the logic a bit. >>>> >>>> This patch combines the walk of the DT nodes with the >>>> code updating the cache size/line_size and nr_sets. >>>> cache_override_properties() (which was DT specific) is >>>> then removed. The result is that cacheinfo.of_node is >>>> no longer used as a temporary place to hold DT references >>>> for future calls that update cache properties. That change >>>> helps to clarify its one remaining use (matching >>>> cacheinfo nodes that represent shared caches) which >>>> will be used by the ACPI/PPTT code in the following patches. >>>> >>>> Cc: Palmer Dabbelt >>>> Cc: Albert Ou >>>> Signed-off-by: Jeremy Linton >>>> --- >>>> ? arch/riscv/kernel/cacheinfo.c |? 1 + >>>> ? drivers/base/cacheinfo.c????? | 65 >>>> +++++++++++++++++++------------------------ >>>> ? include/linux/cacheinfo.h???? |? 1 + >>>> ? 3 files changed, 31 insertions(+), 36 deletions(-) >>>> >>>> diff --git a/arch/riscv/kernel/cacheinfo.c >>>> b/arch/riscv/kernel/cacheinfo.c >>>> index 10ed2749e246..6f4500233cf8 100644 >>>> --- a/arch/riscv/kernel/cacheinfo.c >>>> +++ b/arch/riscv/kernel/cacheinfo.c >>>> @@ -30,6 +30,7 @@ static void ci_leaf_init(struct cacheinfo *this_leaf, >>>> ????????? CACHE_WRITE_BACK >>>> ????????? | CACHE_READ_ALLOCATE >>>> ????????? | CACHE_WRITE_ALLOCATE; >>>> +??? cache_of_set_props(this_leaf, node); >>> >>> This may be necessary but can it be done as later patch ? So far nothing >>> is added that may break riscv IIUC. >> >> Well I think you have a bisection issue where the additional information >> will disappear between this patch and wherever we put this code back in. >> > > Hmm, I am sorry but I fail to see the issue. Before this change, > populate_cache_leaves just populated the info as per ci_leaf_init in > arch/riscv/kernel/cacheinfo.c and cache_override_properties used to fill > the remaining. > > After this patch, the same is achieved in cache_shared_cpu_map_setup. > > In both case, it was by the end of detect_cache_attributes, so I see no > issue. > Hi, I must be misunderstanding something. AFAIK, The code in cache_setup_of_node() won't call cache_of_set_props() because it returns when there is an existing of_node (fw_unique) created by the riscv populate_cache_leaves(). That's why I'm making the direct call here. If we fail to get that change in place before cache_override_properties() is removed then the fields not set by the riscv code (size/etc) will be missing.