On 30/09/2019 11:53:16+0200, Uwe Kleine-König wrote: > Hello, > > On 9/30/19 10:37 AM, Rick Thomas wrote: > > Another case where this argument falls down is the Cubox-iPro which has two > > hardware clocks. The first, labeled /dev/rtc0 by the Debian kernel, is > > accurate as long as the power is up, but it has no battery backup, so when > > the power fails they clock fails as well. The second, labeled /dev/rtc1 by > > Debian, has an optional battery backup, so it can (but may not depending on > > whether the battery is installed) ride thru a power failure. It’s a user > > option whether to install the battery, so again, not something the kernel > > can know at compile time. > > So given that rtc0 is never better than rtc1 it might be beneficial to > remove the rtc0 device there. >
What if rtc0 is the only one able to wake the device or if it has second granularity alarms while rtc1 has minute granularity? -- Alexandre Belloni, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
