Date:   Tue, 24 Jan 2023 13:19:57 -0800
From:   Isaac Manjarres <isaacmanjarres@google.com>
To:     Catalin Marinas <catalin.marinas@arm.com>
Cc:     Andrew Morton <akpm@linux-foundation.org>,
        Saravana Kannan <saravanak@google.com>,
        Suren Baghdasaryan <surenb@google.com>,
        kernel-team@android.com, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v1 0/2] Fixes for kmemleak tracking with CMA regions
Message-ID: <Y9BLfUz5GY69zmjn@google.com>
References: <20230109221624.592315-1-isaacmanjarres@google.com>
 <Y8gpfgEXtialPVLk@arm.com>
 <Y8neaPB2y689WKOf@google.com>
 <Y8/96eIu47UfqsWO@arm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y8/96eIu47UfqsWO@arm.com>
Precedence: bulk

On Tue, Jan 24, 2023 at 03:48:57PM +0000, Catalin Marinas wrote:
> Thanks for digging this out. This patch shouldn't have ended up upstream
> (commit 972fa3a7c17c "mm: kmemleak: alloc gray object for reserved
> region with direct map"). I thought both Calvin Zhang and I agreed that
> it's not the correct approach (not even sure there was a real problem to
> fix).
> 
> Do you still get the any faults with the above commit reverted? I'd
> prefer this if it works rather than adding unnecessary
> kmemleak_alloc/free callbacks that pretty much cancel each-other.
Yes, I still see the same problem after reverting that commit. The problem
still persists because there are CMA areas that are allocated through
memblock_phys_alloc_range(), which invokes kmemleak_alloc_phys(). The
allocation call stack is along the lines of:

kmemleak_alloc_phys()
memblock_alloc_range_nid()
memblock_phys_alloc_range()
__reserved_mem_alloc_size()
fdt_init_reserved_mem()

I also followed up on my suggestion about adding a flags parameter to
the memblock allocation functions to be able to use
MEMBLOCK_ALLOC_NOLEAKTRACE in this particular scenario, but that would
involve changing many call-sites, which doesn't make much sense given
that there are only 4 call-sites that actually use this flag.

Maybe adding a new memblock allocation function that allows this flag to
be passed as just a flag can be used to avoid creating these kmemleak
objects for CMA allocations?

--Isaac