Re: [PATCH v2] writeback: let foreign flushes reach dying cgwbs

[Liz Fong-Jones]

Hi Tejun,

Thank you for your patience with this. It's only my second upstream
contribution to the Linux kernel, and I needed the guidance on how you
handle shared data structures and locking and how you preferred to
resolve this issue.

All addressed in v3:
https://lore.kernel.org/all/20260928-wb-dying-cgwb-flush-v3-0-e35374884667@xxxxxxxxxxxx/

The writeback kick is patch 2. cgwb_create() can run with interrupts
disabled (folio_account_dirtied() -> inode_attach_wb()), so it kicks
the replaced wb after dropping cgwb_lock and makes wb_wakeup()
irq-safe. The cover letter has a test of that case and a question
about the part the kick doesn't cover.

Thanks,
Liz