Date:   Tue, 24 Jan 2023 14:55:41 +0000
From:   Sudeep Holla <sudeep.holla@arm.com>
To:     Conor Dooley <conor.dooley@microchip.com>
Cc:     Geert Uytterhoeven <geert@linux-m68k.org>,
        Pierre Gondois <pierre.gondois@arm.com>,
        linux-kernel@vger.kernel.org, Palmer Dabbelt <palmer@rivosinc.com>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Sudeep Holla <sudeep.holla@arm.com>,
        Will Deacon <will@kernel.org>,
        Paul Walmsley <paul.walmsley@sifive.com>,
        Palmer Dabbelt <palmer@dabbelt.com>,
        Albert Ou <aou@eecs.berkeley.edu>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        Len Brown <lenb@kernel.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Gavin Shan <gshan@redhat.com>,
        linux-arm-kernel@lists.infradead.org,
        linux-riscv@lists.infradead.org, linux-acpi@vger.kernel.org
Subject: Re: [PATCH v4 6/6] arch_topology: Build cacheinfo from primary CPU
Message-ID: <20230124145541.2xwtr7ro2bjnsjd7@bogus>
References: <20230104183033.755668-1-pierre.gondois@arm.com>
 <20230104183033.755668-7-pierre.gondois@arm.com>
 <CAMuHMdUjgxgOXf5He1x=PLn7MQTjZgFQUHj8JrwbyweT4uOALQ@mail.gmail.com>
 <20230124140420.4srnufcvamvff77v@bogus>
 <Y8/tl999NQwbPL/R@wendy>
 <20230124144839.2szjjv256j3pdaif@bogus>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230124144839.2szjjv256j3pdaif@bogus>
Precedence: bulk

On Tue, Jan 24, 2023 at 02:48:39PM +0000, Sudeep Holla wrote:
> 
> Ah, that thread, I remember that :).
> 
> I still need to understand how this is related to memory allocation.
> Pierre was suggesting(in private) if we need to keep fetch_cache_info()
> arch specific but I really don't want to go down that patch until I
> understand and there is no other option.
> 
> Thanks for your time. I will try to recall boot flow and see if I can
> gather the reasoning for the seen behaviour.
> 

OK, I must have atleast taken a look at the code before I replied.
smp_prepare_boot_cpu() is called quite early before page_alloc_init()
and mm_init()(in init_main.c) while smp_prepare_cpus() get called
quite late from kernel_init->kernel_init_freeable().

Geert, can you please try with the patch Conor pointed out and see if
that helps to fix the allocation failures[1]

-- 
Regards,
Sudeep

[1] https://lore.kernel.org/all/20230103035316.3841303-1-leyfoon.tan@starfivetech.com/