Re: [PATCH][RFC] io_uring: Fix io_uring_mmu_get_unmapped_area() for IA-64

On 2023-07-19 17:20, Helge Deller wrote:
> The io_uring testcase is broken on IA-64 since commit d808459b2e31
> ("io_uring: Adjust mapping wrt architecture aliasing requirements").
> 
> The reason is, that this commit introduced an own architecture
> independend get_unmapped_area() search algorithm which doesn't suite 
> the
> memory region requirements for IA-64.
> 
> To avoid similar problems in the future it's better to switch back to
> the architecture-provided get_unmapped_area() function and adjust the
> needed input parameters before the call.  Additionally
> io_uring_mmu_get_unmapped_area() will now become easier to understand
> and maintain.
> 
> This patch has been tested on physical IA-64 and PA-RISC machines,
> without any failures in the io_uring testcases. On PA-RISC the
> LTP mmmap testcases did not report any regressions either.
> 
> I don't expect issues for other architectures, but it would be nice if
> this patch could be tested on other machines too.
> 
> Reported-by: matoro <[email protected]>
> Fixes: d808459b2e31 ("io_uring: Adjust mapping wrt architecture 
> aliasing requirements")
> Signed-off-by: Helge Deller <[email protected]>
> 
> diff --git a/arch/ia64/kernel/sys_ia64.c b/arch/ia64/kernel/sys_ia64.c
> index 6e948d015332..eb561cc93632 100644
> --- a/arch/ia64/kernel/sys_ia64.c
> +++ b/arch/ia64/kernel/sys_ia64.c
> @@ -63,7 +63,7 @@ arch_get_unmapped_area (struct file *filp, unsigned 
> long addr, unsigned long len
>  	info.low_limit = addr;
>  	info.high_limit = TASK_SIZE;
>  	info.align_mask = align_mask;
> -	info.align_offset = 0;
> +	info.align_offset = pgoff << PAGE_SHIFT;
>  	return vm_unmapped_area(&info);
>  }
> 
> diff --git a/arch/parisc/kernel/sys_parisc.c 
> b/arch/parisc/kernel/sys_parisc.c
> index 39acccabf2ed..465b7cb9d44f 100644
> --- a/arch/parisc/kernel/sys_parisc.c
> +++ b/arch/parisc/kernel/sys_parisc.c
> @@ -26,12 +26,17 @@
>  #include <linux/compat.h>
> 
>  /*
> - * Construct an artificial page offset for the mapping based on the 
> physical
> + * Construct an artificial page offset for the mapping based on the 
> virtual
>   * address of the kernel file mapping variable.
> + * If filp is zero the calculated pgoff value aliases the memory of 
> the given
> + * address. This is useful for io_uring where the mapping shall alias 
> a kernel
> + * address and a userspace adress where both the kernel and the 
> userspace
> + * access the same memory region.
>   */
> -#define GET_FILP_PGOFF(filp)		\
> -	(filp ? (((unsigned long) filp->f_mapping) >> 8)	\
> -		 & ((SHM_COLOUR-1) >> PAGE_SHIFT) : 0UL)
> +#define GET_FILP_PGOFF(filp, addr)		\
> +	((filp ? (((unsigned long) filp->f_mapping) >> 8)	\
> +		 & ((SHM_COLOUR-1) >> PAGE_SHIFT) : 0UL)	\
> +	  + (addr >> PAGE_SHIFT))
> 
>  static unsigned long shared_align_offset(unsigned long filp_pgoff,
>  					 unsigned long pgoff)
> @@ -111,7 +116,7 @@ static unsigned long 
> arch_get_unmapped_area_common(struct file *filp,
>  	do_color_align = 0;
>  	if (filp || (flags & MAP_SHARED))
>  		do_color_align = 1;
> -	filp_pgoff = GET_FILP_PGOFF(filp);
> +	filp_pgoff = GET_FILP_PGOFF(filp, addr);
> 
>  	if (flags & MAP_FIXED) {
>  		/* Even MAP_FIXED mappings must reside within TASK_SIZE */
> diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
> index f1b79959d1c1..70eb01faf15f 100644
> --- a/io_uring/io_uring.c
> +++ b/io_uring/io_uring.c
> @@ -3425,8 +3425,6 @@ static unsigned long 
> io_uring_mmu_get_unmapped_area(struct file *filp,
>  			unsigned long addr, unsigned long len,
>  			unsigned long pgoff, unsigned long flags)
>  {
> -	const unsigned long mmap_end = arch_get_mmap_end(addr, len, flags);
> -	struct vm_unmapped_area_info info;
>  	void *ptr;
> 
>  	/*
> @@ -3441,32 +3439,26 @@ static unsigned long 
> io_uring_mmu_get_unmapped_area(struct file *filp,
>  	if (IS_ERR(ptr))
>  		return -ENOMEM;
> 
> -	info.flags = VM_UNMAPPED_AREA_TOPDOWN;
> -	info.length = len;
> -	info.low_limit = max(PAGE_SIZE, mmap_min_addr);
> -	info.high_limit = arch_get_mmap_base(addr, current->mm->mmap_base);
> +	/*
> +	 * Some architectures have strong cache aliasing requirements.
> +	 * For such architectures we need a coherent mapping which aliases
> +	 * kernel memory *and* userspace memory. To achieve that:
> +	 * - use a NULL file pointer to reference physical memory, and
> +	 * - use the kernel virtual address of the shared io_uring context
> +	 *   (instead of the userspace-provided address, which has to be 0UL
> +	 *   anyway).
> +	 * For architectures without such aliasing requirements, the
> +	 * architecture will return any suitable mapping because addr is 0.
> +	 */
> +	filp = NULL;
> +	flags |= MAP_SHARED;
> +	pgoff = 0;	/* has been translated to ptr above */
>  #ifdef SHM_COLOUR
> -	info.align_mask = PAGE_MASK & (SHM_COLOUR - 1UL);
> +	addr = (uintptr_t) ptr;
>  #else
> -	info.align_mask = PAGE_MASK & (SHMLBA - 1UL);
> +	addr = 0UL;
>  #endif
> -	info.align_offset = (unsigned long) ptr;
> -
> -	/*
> -	 * A failed mmap() very likely causes application failure,
> -	 * so fall back to the bottom-up function here. This scenario
> -	 * can happen with large stack limits and large mmap()
> -	 * allocations.
> -	 */
> -	addr = vm_unmapped_area(&info);
> -	if (offset_in_page(addr)) {
> -		info.flags = 0;
> -		info.low_limit = TASK_UNMAPPED_BASE;
> -		info.high_limit = mmap_end;
> -		addr = vm_unmapped_area(&info);
> -	}
> -
> -	return addr;
> +	return current->mm->get_unmapped_area(filp, addr, len, pgoff, flags);
>  }
> 
>  #else /* !CONFIG_MMU */

Tested-by: matoro <[email protected]>

On 6.4.4.  The NFS thing I mentioned in our previous emails appears to 
be a separate regression not related to io_uring.