On Fri, May 30, 2014 at 06:17:49PM +0200, Stephen Warren wrote:

On 05/30/2014 02:23 AM, Peter De Schrijver wrote:

quoted

On Thu, May 29, 2014 at 09:01:27PM +0200, Stephen Warren wrote:

quoted

On 05/28/2014 06:54 AM, Peter De Schrijver wrote:

quoted

This driver allows userspace to read the raw efuse data. Its userspace
interface is modelled after the sunxi_sid driver which provides similar
functionality for some Allwinner SoCs. It has been tested on
Tegra20 (ventana), Tegra30 (beaverboard), Tegra114 (dalmore) and
Tegra124 (jetson TK1).

quoted

Changes since v4:

* Provide fallback to hardcoded 0x70000800 in case the apbmisc DT node is
  missing. This is exactly what the current code does and prevents a system
  crash in that case due to an invalid memory access by tegra_read_chipid()

Wouldn't it be better to simply return an error?

This would mean you can't boot a system with these patches applied unless you
also update the device tree. The system would crash during boot because CCF
relies on tegra_read_chipid() as an APB barrier. Also tegra_boot_secondary()
relies on the chipid to select the correct method for booting secondary cores.

Is this series really backwards-compatible anyway? tegra_init_fuse()
contains a whole bunch of places where resources are pulled out of DT,
and only one of those has a fallback to use APBMISC_BASE if the DT entry
isn't present.

Apart from the randomness initialization, there are no users of
tegra_read_straps(). tegra_fuse_readl() is a new API, so that doesn't pose
any backwards-compatibility problems. The only bug I noticed when rereading
the code is that tegra_get_revision() reads the fuses on an A03 Tegra20 to
findout if it's A03 or A03p. This would obviously crash if the DT entries
are missing. I think this can be worked around by claiming it's an A03
in case the DT entries are missing. I don't think that will cause any problems?

Cheers,

Peter.