linux-kernel - Re: [PATCH] PCI: mvebu: Fix the use of the for_each_of

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 for Android: free password hash cracker in your pocket

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

Message-ID: <34a23816bf125c577a03ba68f251e7a77db9805b.camel@gmail.com>
Date: Sun, 07 Sep 2025 11:28:04 +0200
From: Klaus Kudielka <klaus.kudielka@...il.com>
To: Jan Palus <jpalus@...tmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@...tlin.com>, Pali
 Rohár	 <pali@...nel.org>, Lorenzo Pieralisi
 <lpieralisi@...nel.org>, Krzysztof Wilczyński	
 <kwilczynski@...nel.org>, Manivannan Sadhasivam <mani@...nel.org>, Rob
 Herring	 <robh@...nel.org>, Bjorn Helgaas <bhelgaas@...gle.com>, 
	linux-pci@...r.kernel.org, linux-arm-kernel@...ts.infradead.org, 
	linux-kernel@...r.kernel.org, Bjorn Helgaas <helgaas@...nel.org>
Subject: Re: [PATCH] PCI: mvebu: Fix the use of the for_each_of_range()
 iterator

On Wed, 2025-09-03 at 14:44 +0200, Jan Palus wrote:
> 
> Thanks for the patch Klaus! While it does improve situation we're not
> quite there yet. It appears that what used to be stored in `cpuaddr` var
> is also very different from `range.cpu_addr` value so the results
> in both `*tgt` and `*attr` are both wrong.
> 
> Previously `cpuaddr` had a value like ie 0x8e8000000000000 or
> 0x4d0000000000000. Now `range.cpu_addr` is always 0xffffffffffffffff.
> Luckily what used to be stored in `cpuaddr`:
> 
> u64 cpuaddr = of_read_number(range + na, pna)
> 
> appears to be stored in range.pci_bus_addr now. I can't make any
> informed comment about this discrepancy however I can confirm following
> change (in addition to your patch) makes mvebu driver work again (or at
> least like it used to work in 6.15, it still needs Pali's patches to
> have some devices working):
> 
> -			*tgt = DT_CPUADDR_TO_TARGET(range.cpu_addr);
> -			*attr = DT_CPUADDR_TO_ATTR(range.cpu_addr);
> +			*tgt = DT_CPUADDR_TO_TARGET(range.parent_bus_addr);
> +			*attr = DT_CPUADDR_TO_ATTR(range.parent_bus_addr);


Oh well, a CPU address that is not a CPU address....

Looking again at of_range_parser_one() - which I assume to be "correct" here:

	range->flags = parser->bus->get_flags(parser->range);

	range->bus_addr = of_read_number(parser->range + busflag_na, na - busflag_na);

	if (parser->dma)
		range->cpu_addr = of_translate_dma_address(parser->node,
				parser->range + na);
	else
		range->cpu_addr = of_translate_address(parser->node,
				parser->range + na);

	range->parent_bus_addr = of_read_number(parser->range + na, parser->pna);

Indeed, range.cpu_addr is a translated version of range.parent_bus_addr, and does
not seem to be relevant to pci-mvebu.

The driver previously interpreted the raw part of the "/soc/pcie/ranges" resource
(storing it in a local variable called cpuaddr).

To restore the behavior completely, we thus can use range.parent_bus_addr ---
confirming your test.

I'll prepare a patch version 2, including this second fix.