create an autotest to ensure reading and writing to vpd works ok |
|
Issue descriptionRequest from stephenlin@ I think it would be useful to have a test to determine if there are any reliability problems in read the RW_VPD and writing to it. We have seen problems writing to the vpd on some DUTs on rlz_CheckPing https://stainless.corp.google.com/search?view=list&first_date=2018-12-05&last_date=2018-12-06&test=%5Erlz%5C_CheckPing%24&build=R73-*&status=FAIL&status=ERROR&exclude_cts=true&exclude_not_run=false&exclude_non_release=true&exclude_au=true&exclude_acts=true&exclude_retried=true&exclude_non_production=false
,
Dec 14
- Are there specific values in the VPD that are more prone to errors than others? Or should it not matter If there are then I'm not aware. I don't think the code tries to read/write VPD values very often. That is probably somewhat unique to should_send_rlz_ping and first_active_omaha_ping_sent. The u_bind_attribute for ECHO (offers) is read when the user tries to redeem an offer; that shouldn't happen too often, but there are occasional support issues with unknown causes where a user cannot redeem an offer which could be related to a failure to read the RW_VPD. - Which VPD values specifically would you like added to the test? first_active_omaha_ping_sent should_send_rlz_ping Can we test both reading/writing? I really don't think it matters what the attribute is assuming there is an issue at all. - Do we need to reboot at any point? No. This test would be testing the raw read/write ability for the RW_VPD. There are layers above that which have caching, but we're not testing that at all. - How many times should the test read/write to VPD? Not sure, but as it should be pretty quick and this is not necessarily a common problem, then whatever can fit into a relatively short testing time. - Previously I have added this line to tests that do any VPD stuff "dump_vpd_log --force". How necessary is this? That's not necessary if you're testing the lowest level of read/write. The dump_vpd_log --force is needed to flush cached values, but you're reading straight from the RW_VPD so the cache isn't relevant. - Any other pointers you want to give here Assuming there is some flakiness in the reading/writing of the RW_VPD I wouldn't want just pass/fail, but to understand just how flaky the reading/writing is. As such, the test should ideally do like 100+ cycles of writing/reading. |
|
►
Sign in to add a comment |
|
Comment 1 by dhaddock@google.com
, Dec 6Owner: stephenlin@chromium.org