Bug 232976 – Log memory usage metadata when WebContent crosses critical or warning level memory thresholds

• Status: RESOLVED
• Resolution: FIXED
• Component: WebCore Misc.

[Ben Nham]

We receive a large number of out of memory reports without enough logging info to properly debug. We should log more metadata about memory usage (similar to what we do on Mac in `WebCore::logMemoryStatisticsAtTimeOfDeath`) to help us debug those reports.

[Ben Nham]

Created attachment 443891 [details]
Patch

[Ben Nham]

<rdar://85171163>

[Ben Nham]

Created attachment 443894 [details]
Patch

[Darin Adler]

Comment on attachment 443894 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=443894&action=review

> Source/WTF/wtf/cocoa/ResourceUsageCocoa.cpp:140
> +            // Count all resident malloc pages as dirty.
> +            size_t dirtyPages = info.pages_resident - info.pages_reusable;
> +            tags[info.user_tag].dirty += dirtyPages + info.pages_swapped_out;

Do we really need a local variable just to add pages_swapped_out?

> Source/WTF/wtf/cocoa/ResourceUsageCocoa.cpp:143
> +            tags[info.user_tag].dirty += (info.pages_dirtied + info.pages_swapped_out);

Not sure the parentheses add clarity here.

> Source/WebCore/page/MemoryRelease.h:34
> +enum class LogMemoryStatisticsReason {

: uint8_t?

> Source/WebKit/WebProcess/WebProcess.cpp:464
> +                // Log stats in the next turn of the run loop so that it runs after the low memory handler.

This comment is here, but not above. Does the same rationale apply above?

[Ben Nham]

Created attachment 443977 [details]
patch for landing

[Ben Nham]

Created attachment 443979 [details]
patch for landing

[EWS]

Committed r285648 (244148@main): <https://commits.webkit.org/244148@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 443979 [details].