armv8rng: Fix an inverted test in random_rndr_read_one()

[markj]

Summary

If we get a random number, the NZCV is set to 0b0000. Then
"cset %w1, ne" will test whether Z == 0 and set %w1 to 1 if so.
More specifically, "cset %w1, ne" maps to "csinc %w1, wzr, wzr, eq",
which stores 0 in %w1 when NZCV == 0b0100 and 1 otherwise.

Thus, on a successful read we expect ret != 0, so the loop condition
needs to be fixed.

Fixes: 9eecef052155 ("Add an Armv8 rndr random number provider")
Reported by: Kevin Day <[e-mail]>

Test Plan

I have no hardware to test this with.

[jrtc27]

So the current code tries 10 times to hit a failure, and will use the last value if all 10 reads succeeded, given the condition after the loop is correct? Importantly, there's no security issue here, because we didn't actually report a successful read when it failed?

[markj]

In D54259#1239893, @jrtc27 wrote:

So the current code tries 10 times to hit a failure, and will use the last value if all 10 reads succeeded, given the condition after the loop is correct? Importantly, there's no security issue here, because we didn't actually report a successful read when it failed?

Yes, I think that's right. At worst we are not getting entropy because reading the register 10 times back-to-back causes a failure. Our default CSPRNG by design draws entropy from many different sources, so this isn't catastrophic.