On 18.01.24 14:35, Yang Shi wrote:
> From: Yang Shi <yang@xxxxxxxxxxxxxxxxxxxxxx>
>
> The commit efa7df3e3bb5 ("mm: align larger anonymous mappings on THP
> boundaries") caused two issues [1] [2] reported on 32 bit system or compat
> userspace.
>
> It doesn't make too much sense to force huge page alignment on 32 bit
> system due to the constrained virtual address space.
>
> [1] https://lore.kernel.org/linux-mm/CAHbLzkqa1SCBA10yjWTtA2mKCsoK5+M1BthSDL8ROvUq2XxZMw@xxxxxxxxxxxxxx/T/#mf211643a0427f8d6495b5b53f8132f453d60ab95
> [2] https://lore.kernel.org/linux-mm/CAHbLzkqa1SCBA10yjWTtA2mKCsoK5+M1BthSDL8ROvUq2XxZMw@xxxxxxxxxxxxxx/T/#me93dff2ccbd9902c3e395e1c022fb454e48ecb1d
[FWIW, this is now 4ef9ad19e17676 ("mm: huge_memory: don't force huge
page alignment on 32 bit") in mainline]
Quick question: it it okay to ask Greg to pick this up for linux-6.7.y
series?
I'm wondering because Jiri's report ([1] in above quote) sounded like
this is something that will affect and annoy quite a few people with the
linux-6.7.y.
Ciao, Thorsten
> Fixes: efa7df3e3bb5 ("mm: align larger anonymous mappings on THP boundaries")
> Reported-by: Jiri Slaby <jirislaby@xxxxxxxxxx>
> Reported-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
> Tested-by: Jiri Slaby <jirislaby@xxxxxxxxxx>
> Tested-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
> Cc: Rik van Riel <riel@xxxxxxxxxxx>
> Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx>
> Cc: Christopher Lameter <cl@xxxxxxxxx>
> Signed-off-by: Yang Shi <yang@xxxxxxxxxxxxxxxxxxxxxx>
> ---
> mm/huge_memory.c | 9 +++++++++
> 1 file changed, 9 insertions(+)
>
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 94ef5c02b459..e9fbaccbe0c0 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -37,6 +37,7 @@
> #include <linux/page_owner.h>
> #include <linux/sched/sysctl.h>
> #include <linux/memory-tiers.h>
> +#include <linux/compat.h>
>
> #include <asm/tlb.h>
> #include <asm/pgalloc.h>
> @@ -811,6 +812,14 @@ static unsigned long __thp_get_unmapped_area(struct file *filp,
> loff_t off_align = round_up(off, size);
> unsigned long len_pad, ret;
>
> + /*
> + * It doesn't make too much sense to froce huge page alignment on
> + * 32 bit system or compat userspace due to the contrained virtual
> + * address space and address entropy.
> + */
> + if (IS_ENABLED(CONFIG_32BIT) || in_compat_syscall())
> + return 0;
> +
> if (off_end <= off_align || (off_end - off_align) < size)
> return 0;
>
