As I mentioned in a recent post on this thread, I updated firmware on a Samsung 980 Pro to the latest, released within about the last month or so, and my a Monterey boot stalls for several minutes. The console output is interrupted mid line on a message unrelated to Trim, but for the duration of the stall, the board I/O LED flashes with heavy continuous activity.
From this I want to speculate from others' boot console output that Monterey issues Trim for all partition free space and that Samsung firmware update has changed Trim command semantics in a way such that what was either a very quick operation, internal background operation — or no operation! — is now synchronous WRT to the storage driver.
For example, maybe...
- ...Trim went from being advisory, something the drive controller skipped if it was busy — or if the device slippery just wanted to ignore it — to mandatory?
- ...there's an evolving interpretation that Trim blocks must be in some specific state of hygiene (e.g., zero) when the Trim command completes, and to achieve this the drive must block until the literal garbage collection gets done?
- ...Apple NVMe devices have special semantics that Monterey support depends upon, such as non-blocking Trim, and Apple moved Trim from a chore that was previously limited to SW updates or fsck to a routine boot activity, keeping in mind that Mac is going the way of iPhone as a device with a power envelope that allows expectation that the user will rarely restart it, and only do so during an activity which is expected to take a long time, such as a SW update.
... Etc.
I don't know if any of this is true but I suggest it as a way to think about the situation.
My thoughts keep returning to problems with Sabrent Rocket 4 deaths: These happened under Big Sur. Maybe Big Sur was mixing Trim with other I/O on the fly and there's a pathological interaction on Phison E15 rev x.y.z?
Other though is how does Trim operating on Flash pages interact with LBA block-level data structures on the drive. Is the OS supposed to know not to Trim less than flash-page units? What does macOS think it's safe to Trim and how does that relate to the drive layout?
...Etc, etc.
These thoughts suggest the challenges that will make hackintosh problematic going forward. What we should note is that you can't approach overcoming such challenges from a point of view of "know to work" list, because the set of what is "known to work" intersects with "that which has not failed -yet-"
In the classic 90's Stephen Frears movie High Fidelity, young John Cusack asks an even younger Jack Black: "If I told you I hadn't seen the sequel to Evil Dead 'yet', would you think that I fully intended to see the movie?" He's wondering out loud because his newly estranged girlfriend just retorted to his query as to whether she sleeping with other guys with a blithe 'not yet'. In the movie, Cusack's character has wild fantasies about what his ex-GF is doing that culminate in a showdown with a patchouli-scented free-love guru and a old-school Bell telephone.
My speculations about Trim are like this movie character.
Where we can keep an eye out is for insights about the state of Trim in the industry and trends, and instrumentation that let's us see a bit more about what's happening under the covers of macOS.
