`pr_cont`: consider removing it

[ojeda]

@nbdd0121

It seems that KERN_CONT is heavily discouraged: the comment in kern_levels.h says:

Only to be used by core/arch code during early bootup (a continued line is not SMP-safe otherwise).

Maybe we shouldn't expose it in Rust at all? With powerful formatting capability of core::fmt I don't see it being very useful anyway.

@ojeda

Agreed, even checkpatch.pl warns about it. And yet there are ~1k uses inside drivers/... :/ When I wrote this I was trying to produce the same functionality, more or less. But yeah, removing would simplify a few conditionals and one macro case.

---

Sample a few C drivers where pr_cont is used, and try to see what are their use cases. Then, taking that into account, decide and justify whether we should remove it from rust/kernel/print.rs (e.g. if it looks like all the C use cases can be easily done with Rust's formatting, or if it looks most of them should not be using pr_cont(), etc.).

[obviyus]

Some notes after perusing a couple of C drivers:

• pr_cont() seems to be commonly used as a one-word follow-up for a previous pr_* invocation. I haven't yet seen it used with substitutions. Seems it's used quite frequently with substitutions as well: see L238/239 of nxt6000.c.
• All invocations of multiple pr_cont()s invariably end with a pr_cont("\n"); at the end.
• For the majority of cases, pr_cont() just seems to be used as a one-off print appending the previous one. It seems like it should be quite trivial to replace pr_cont() invocations with a single call to any of the pr_*! macros using format!() or simple substitutions.
• However, in cases like L298 of ssb/pci.c and L1707 of floppy.c, pr_cont() is used to show progress of some execution as it occurs during a loop. I'm not sure if this is possible to achieve without pr_cont().
• Another outlier is L1382 of floppy.c where pr_cont() is called inside the tell_sector() function in continuation to a DPRINT(). Should be possible to have tell_sector() return the string in question to the invoking expression and store it for a single pr_* invocation.

Apologies if I misunderstood anything. Please let me know if I should sample a couple more drivers.

[nbdd0121]

L1707 of floppy.c

This isn't actually showing progress; it's only printing an array.

L298 of ssb/pci.c

This looks like the only case where pr_cont usage is inevitable, aka showing progress. But that could probably be replaced with multiple lines.