On Tue, May 03, 2022 at 09:49:36AM +0200, Daniels Umanovskis wrote: > On 5/3/22 2:20 AM, Rob Herring wrote: > > My first question is whether this is board specific? If the default or > > what the reference manual says is 500us, but you have a case needing > > 600us, why not just change the driver. I don't think this really needs > > tuning to each board unless the delay becomes noticeable. > > This isn't a board specific issue. I detected the issue on a board we've > been using for a long time, it's due to a specific batch of USB2512 hubs. > We've had several batches of these hubs that were fine, but more recently > received a batch produced in late 2021 and the hubs in this batch do not > become responsive to I2C within the expected 500us. What's the maximum timeout you've observed? I guess it would be the simpler and "better" approach to just increase the timeout in the driver (if it's not too much above the 500µs). > > I arrived at that by using an oscilloscope to observe how soon after > deasserting the reset signal the USB hub is able to respond to I2C. Most of > the 2512s we have do that within 500us, the latest batch doesn't. > regards;rl