ringbuf: rewrite entry to have no panics

[cbiffle]

This knocks about 120 bytes off of each ringbuf in a task (because the Ringbuf::entry routine gets specialized for each ringbuf, and this makes that routine smaller).

Reduces Gimlet flash image size by 3.4%.

[hawkw]

This is lovely, and I really appreciate the comments (as usual!)

[hawkw]

Hmm, I have some bad news about this. After rebasing my branch from #1621 to use this code for Ringbuf::entry rather than the code on master, all my ringbufs appear to be empty:

eliza@cadbury ~ $ pfexec humility -t c71 ringbuf
humility: attached to 0483:3754:001100184741500820383733 via ST-Link V3
humility: ring buffer drv_gimlet_seq_server::__RINGBUF in gimlet_seq:
humility: ringbuf dump failed: member missing from struct: cell
humility: ring buffer drv_i2c_devices::adm1272::__RINGBUF in power:
humility: ring buffer drv_i2c_devices::max31790::__RINGBUF in thermal:
humility: ring buffer drv_i2c_devices::sbrmi::__RINGBUF in sbrmi:
humility: ring buffer drv_oxide_vpd::__RINGBUF in gimlet_seq:
humility: ring buffer drv_oxide_vpd::__RINGBUF in host_sp_comms:
humility: ring buffer drv_packrat_vpd_loader::__RINGBUF in gimlet_seq:
humility: ring buffer drv_sbrmi::__RINGBUF in sbrmi:
humility: ring buffer drv_stm32h7_spi_server_core::__RINGBUF in spi2_driver:
humility: ring buffer drv_stm32h7_spi_server_core::__RINGBUF in sprot:
humility: ring buffer drv_stm32h7_sprot_server::__RINGBUF in sprot:
humility: ring buffer drv_stm32xx_i2c::__RINGBUF in i2c_driver:
humility: ring buffer drv_stm32xx_i2c::__RINGBUF in spd:
humility: ring buffer drv_stm32xx_i2c_server::__RINGBUF in i2c_driver:
humility: ring buffer ksz8463::__RINGBUF in net:
humility: ring buffer ksz8463::__RINGBUF in host_sp_comms:
humility: ring buffer stm32h7_update_server::__RINGBUF in update_server:
humility: ring buffer task_control_plane_agent::CRITICAL in control_plane_agent:
humility: ring buffer task_control_plane_agent::__RINGBUF in control_plane_agent:
humility: ring buffer task_control_plane_agent::update::rot::__RINGBUF in control_plane_agent:
humility: ring buffer task_dump_agent::__RINGBUF in dump_agent:
humility: ring buffer task_dump_agent::udp::__RINGBUF in dump_agent:
humility: ring buffer task_hiffy::stm32h7::__RINGBUF in hiffy:
humility: ring buffer task_host_sp_comms::__RINGBUF in host_sp_comms:
humility: ring buffer task_jefe::dump::__RINGBUF in jefe:
humility: ring buffer task_jefe::external::__RINGBUF in jefe:
humility: ring buffer task_net::mgmt::__RINGBUF in net:
humility: ring buffer task_packrat::__RINGBUF in packrat:
humility: ring buffer task_power::__RINGBUF in power:
humility: ring buffer task_power::bsp::__RINGBUF in power:
humility: ring buffer task_spd::__RINGBUF in spd:
humility: ring buffer task_thermal::__RINGBUF in thermal:
humility: ring buffer task_validate::__RINGBUF in validate:
humility: ring buffer vsc85xx::__RINGBUF in net:

After changing my PR back to use the previous code, the ringbufs contain entries as expected:

eliza@cadbury ~ $ pfexec humility -t c71 ringbuf
humility: attached to 0483:3754:001100184741500820383733 via ST-Link V3
humility: ring buffer drv_gimlet_seq_server::__RINGBUF in gimlet_seq:
humility: ringbuf dump failed: member missing from struct: cell
humility: ring buffer drv_i2c_devices::adm1272::__RINGBUF in power:
 NDX LINE      GEN    COUNT PAYLOAD
   0  106        1        1 Config(CommandData(0x3f3f))
   1  159        1        1 Coefficients(Coefficients { m: 0xfde, b: 0x0, R: 0xfe })
   2  177        1        1 Coefficients(Coefficients { m: 0x297, b: 0x5000, R: 0xff })
   3  205        1        1 Coefficients(Coefficients { m: 0x2927, b: 0x0, R: 0xfd })
   4  106        1        1 Config(CommandData(0x3f3f))
   5  159        1        1 Coefficients(Coefficients { m: 0xfde, b: 0x0, R: 0xfe })
   6  177        1        1 Coefficients(Coefficients { m: 0x52e, b: 0x5000, R: 0xff })
   7  205        1        1 Coefficients(Coefficients { m: 0x524e, b: 0x0, R: 0xfd })
humility: ring buffer drv_i2c_devices::max31790::__RINGBUF in thermal:
humility: ring buffer drv_i2c_devices::sbrmi::__RINGBUF in sbrmi:
humility: ring buffer drv_oxide_vpd::__RINGBUF in gimlet_seq:
humility: ring buffer drv_oxide_vpd::__RINGBUF in host_sp_comms:
humility: ring buffer drv_packrat_vpd_loader::__RINGBUF in gimlet_seq:
humility: ring buffer drv_sbrmi::__RINGBUF in sbrmi:
humility: ring buffer drv_stm32h7_spi_server_core::__RINGBUF in spi2_driver:
 NDX LINE      GEN    COUNT PAYLOAD
  52  494     5450        1 WaitISR(0x32017)
  53  464     5450        1 Rx(0x0)
  54  464     5450        1 Rx(0x7)
  55  494     5450        1 WaitISR(0x10012)
  56  464     5450        1 Rx(0x6)
  57  322     5450        1 Start(exchange, (0x2, 0x4))

 < ... REDACTED FOR BREVITY ... >

(n.b. that the gimlet_seq ringbuf parse failure is because i'm using that ringbuf for testing my event counts change and should be expected).

I haven't tested just this branch yet, so it's possible that it's an interaction with my code, but I intend to shortly --- @cbiffle, have you actually tried this out yet?

Edit: Disregard all this, I messed up my rebase and ate the line that actually updates the ringbuf's last element. Whoopsie.

[cbiffle]

Ringbufs with this change check out locally; issue appears to have been a conflict between our changes.

[dancrossnyc]

Thanks for the comment! I'd remove this bullet point, though, as this code would be correct regardless. Specifically, one presumes that with modulus, this code would be written as let ndx = last.wrapping_add(1) % self.buffer.len();. Since usize::MAX.wrapping_add(1) is 0, and 0 (mod N) is 0 for N != 0, this would still (correctly) be 0.