On Wed, Nov 20, 2024 at 05:37:35PM +0100, Caleb Connolly wrote:
On 20/11/2024 17:03, Simon Glass wrote:
Hi Caleb,
On Wed, 20 Nov 2024 at 08:56, Caleb Connolly caleb.connolly@linaro.org wrote:
Hi Simon,
On 20/11/2024 16:35, Simon Glass wrote:
This series provides a way to tell a serial UART that it can't actually work, at runtime. The main motivation is to deal with a coreboot feature where it does not provide UART details in the sysinfo structure unless the UART is also enabled in coreboot.
Why is the UART driver probed if coreboot doesn't provide the necessary info for it?
U-Boot doesn't do well without some sort of serial device. That could perhaps be improved. But the point here is that we need a single U-Boot build which can start from coreboot, regardless of what coreboot decides to do.
Right, I'm suggesting that at runtime you notice that coreboot didn't provide a serial port definition (or gave you a bogus address or something) and skip probing the serial port in that case.
Couldn't you disable CONFIG_REQUIRE_SERIAL_CONSOLE and skip probing it?
No, because I want it to work if the serial info is available.
This option doesn't disable the serial port, it just makes U-Boot not panic if it can't find one. For example I enable this when building U-Boot for the upstream supported Qualcomm phones since many don't enable the serial port in their DTS.
Yes, this feels like an oddly coreboot specific solution to a generic problem that we should be able to handle already I thought? Like a Pi can have or not have a serial port enabled and still have video and we use that. We might well need something specific to know the port is disabled since I gather we aren't getting a devicetree that says status = disabled in it. But this ought to be a common enough general case?