On Tue, 11 Feb 2020, Martin Townsend wrote: > On Mon, Feb 10, 2020 at 3:12 PM Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > > > > On Mon, 10 Feb 2020, Martin Townsend wrote: > > > > > Hi, > > > > > > We are using the USB mass storage gadget driver in Linux and > > > everything is working fairly well but sometimes we are finding that we > > > are exceeding the 100mA limit (which I think is the default the host > > > will provide) before Linux has had a change to enumerate the USB > > > device at which point we ask for 500mA. We have tried to reduce the > > > power by ensuring all clocks are turned off, devices not used disabled > > > etc but can't seem to satisfy the 100mA limit. We were thinking that > > > maybe we could enumerate the USB device during U-Boot at which point > > > we know we are under the 100mA limit. Does anyone know a reason why > > > this would not work? > > > > It won't work because U-Boot isn't the kernel. When the kernel takes > > over the USB controller, it will force the host to re-enumerate the > > gadget -- and you will be right back where you started. > > > > The only way to make this work would be to prevent U-Boot from ever > > actually booting the Linux kernel. This probably isn't the way you > > want your device to behave, though. > > > > > would at some point from U-Boot enumerating a new > > > bMaxPower would the connection be reset and drop back to 100mA during > > > the kernel startup? Are there any considerations that we would need to > > > take care of in the kernel for this to work? > > > > If you can't satisfy the 100-mA current limit then you should make your > > device be self-powered instead of relying on power delivered over the > > USB bus. > > > > Be aware also that bus-powered USB-2 hubs can't deliver 500 mA to their > > downstream ports. They are always limited to 100 mA or less. If your > > gadget was plugged into one of them, it wouldn't be able to work at > > all. > > > > Alan Stern > > > > Alan, > Thank you for the swift response. On the first point, if we modify > the kernel to re-enumerate the same bMaxPower of 500mA would this > still be a problem? ie is there a period when the kernel takes over > the USB controller that the Host side would go back to 100mA after > U-Boot negotiates the 500mA? Yes, there is. Remember, the kernel knows nothing of what U-Boot has done. It has to start fresh. > The last point you make is a very good > one and something I hadn't thought of. If we can overcome the first > hurdle I think we would be happy for it to be a requirement for the > device to be plugged directly into a PC. Or into a powered hub. Alan Stern
