Re: [PATCH v30 2/6] fs/proc/task_mmu: Implement IOCTL to get and optionally clear info about PTEs

[Muhammad Usama Anjum]

On 8/16/23 2:45 PM, Michał Mirosław wrote:
> On Wed, Aug 16, 2023 at 11:59:21AM +0500, Muhammad Usama Anjum wrote:
>> The PAGEMAP_SCAN IOCTL on the pagemap file can be used to get or optionally
>> clear the info about page table entries.
> [...]
>> --- a/fs/proc/task_mmu.c
>> +++ b/fs/proc/task_mmu.c
> [...]
>> +static long do_pagemap_scan(struct mm_struct *mm, unsigned long uarg)
>> +{
> [...]
>> +	for (walk_start = p.arg.start; walk_start < p.arg.end;
>> +			walk_start = p.arg.walk_end) {
>> +		long n_out;
>> +
>> +		if (fatal_signal_pending(current)) {
>> +			ret = -EINTR;
>> +			break;
>> +		}
>> +
>> +		ret = mmap_read_lock_killable(mm);
>> +		if (ret)
>> +			break;
>> +		ret = walk_page_range(mm, walk_start, p.arg.end,
>> +				      &pagemap_scan_ops, &p);
>> +		mmap_read_unlock(mm);
>> +
>> +		n_out = pagemap_scan_flush_buffer(&p);
>> +		if (n_out < 0)
>> +			ret = n_out;
>> +		else
>> +			n_ranges_out += n_out;
>> +
>> +		p.arg.walk_end = p.walk_end_addr ? p.walk_end_addr : p.arg.end;
> 
> I think p.walk_end_addr can be removed and replaced by `p.arg.walk_end`
> directly in the walk functions. If we don't set walk_end_addr we'll also
> return 0 so the check below will match. Might be good to add this as
> a comment.
I'll remove it and add a short comment.

> 
>> +		if (ret != -ENOSPC)
>> +			break;
>> +
>> +		if (p.arg.vec_len == 0 || p.found_pages == p.arg.max_pages)
>> +			break;
>> +	}
>> +
>> +	/* ENOSPC signifies early stop (buffer full) from the walk. */
>> +	if (!ret || ret == -ENOSPC)
>> +		ret = n_ranges_out;
>> +
>> +	p.arg.walk_end = p.arg.walk_end ? p.arg.walk_end : walk_start;
> 
> When the walk is finished, with ret == 0, the walk_start will point to
> the beginning, not the end of the range. So:
> 
> if (!walk_end) walk_end = p.arg.end;
This condition is to cater for the case when for loop doesn't execute at
all because the address range was zero. In that case start == end. So
p.arg.start or p.arg.end both would work fine. I'll add p.arg.end in
accordance to above loop.

> 
> Other than that, the patch looks complete now. Thanks for all your work!
I'll send the next revision.

> 
> Best Regards
> Michał Mirosław

-- 
BR,
Muhammad Usama Anjum