This patchset contains two small patches. Identified this when I tried to fix a kdump bug that page allocation failure is always warned out during kdump kernel on x86_64. Since dma atomic pool is initialized very early during bootup, that allocation failure is always triggered in dma_atomic_pool_init() firstly. I originally planned to add an kernel option to disable each of the three dma atomic pool. Finally I realized it's not only dma atomic pool getting the issue, any later requsting page from DMA zone will fail. So I changed solution to fix the kdump bug. While the document clean up and the small fix which allows user to disable atomic pool makes sense to improve. The small fix which makes 'coherent_pool=0' disable dma atomic pool like 'cma=0', looks more reasonable than the old behavioud which will take the default atomic size. [PATCH RESEND v2 0/5] Avoid requesting page from DMA zone when no managed pages https://lore.kernel.org/all/20211207030750.30824-1-bhe@xxxxxxxxxx/T/#u Baoquan He (2): docs: kernel-parameters: Update to reflect the current default size of atomic pool dma-pool: allow user to disable atomic pool Documentation/admin-guide/kernel-parameters.txt | 5 ++++- kernel/dma/pool.c | 7 +++++-- 2 files changed, 9 insertions(+), 3 deletions(-) -- 2.26.0
