On Mon, Oct 13, 2025 at 06:55:00AM +0530, Deepanshu Kartikey wrote:

Hi Sasha,

Please do NOT backport commit dd83609b8898 alone to stable. This patch causes a regression in fallocate(PUNCH_HOLE) operations where pages are not freed immediately, as reported by Mark Brown.

The fix for this regression is already in linux-next as commit 91a830422707 ("hugetlbfs: check for shareable lock before calling huge_pmd_unshare()").

Please backport both commits together to avoid introducing the regression in stable kernels:

• dd83609b88986f4add37c0871c3434310652ebd5 ("hugetlbfs: skip VMAs without shareable locks in hugetlb_vmdelete_list")
• 91a830422707a62629fc4fbf8cdc3c8acf56ca64 ("hugetlbfs: check for shareable lock before calling huge_pmd_unshare()")

As this is not in linux-next yet, I'll just drop this original patch for now. When it does land in Linus's tree, please let us know so we can submit both of them to the stable queues.

thanks,

greg k-h