linux-kernel - remoteproc: rproc_va_to_pa returns invalid physical address when using sparse memory model

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <20211108094808.1993-1-mark-pk.tsai@mediatek.com>
Date:   Mon, 8 Nov 2021 17:48:08 +0800
From:   Mark-PK Tsai <mark-pk.tsai@...iatek.com>
To:     <ohad@...ery.com>, <bjorn.andersson@...aro.org>,
        <mathieu.poirier@...aro.org>
CC:     <linux-remoteproc@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <mark-pk.tsai@...iatek.com>, <yj.chiang@...iatek.com>
Subject: remoteproc: rproc_va_to_pa returns invalid physical address when using sparse memory model

On my arm platform with CONFIG_SPARSEMEM=y, when the virtual address comes
from ioremap, which map to a reserved memory region, rproc_va_to_pa
returns a invalid address. (no iommu)

It's because the corresponding struct page and section not present.
And then __page_to_pfn read the page->flags in the returned page which
is actually an invalid address.
(When CONFIG_SPARSMEM=y && CONFIG_SPARSEMEM_VMEMMAP=n, kernel get the
section the page belong to in flags field.)

I'm looking for suggestion to properly fix this problem.
Could you please give us some suggestion?