On Thu, Dec 05, 2024 at 01:53:29PM -0800, Gwendal Grignou wrote: > Since limiting the queue depth to 2 is only needed for a small subset > of eMMC memory modules that can be connected behind the bridge, would > it make sense to apply this patch, but add the kernel module parameter > mentioned earlier for impacted devices? I don't think qd2 is needed for anyone, though. The problem sounds like it was root caused to a boundary condition, which is addressed in a different patch. But I am not 100% sure, as the thread was left hanging for some external confirmation: https://lore.kernel.org/linux-nvme/Z0DdU9K9QMFxBIL8@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/