Re: [PATCH 11/12] device-dax: Add dis-contiguous resource support

From: Joao Martins
Date: Wed Mar 25 2020 - 06:36:50 EST


On 3/24/20 4:12 PM, Joao Martins wrote:
> On 3/23/20 11:55 PM, Dan Williams wrote:
>> static ssize_t dev_dax_resize(struct dax_region *dax_region,
>> struct dev_dax *dev_dax, resource_size_t size)
>> {
>> resource_size_t avail = dax_region_avail_size(dax_region), to_alloc;
>> - resource_size_t dev_size = range_len(&dev_dax->range);
>> + resource_size_t dev_size = dev_dax_size(dev_dax);
>> struct resource *region_res = &dax_region->res;
>> struct device *dev = &dev_dax->dev;
>> - const char *name = dev_name(dev);
>> struct resource *res, *first;
>> + resource_size_t alloc = 0;
>> + int rc;
>>
>> if (dev->driver)
>> return -EBUSY;
>> @@ -684,38 +766,47 @@ static ssize_t dev_dax_resize(struct dax_region *dax_region,
>> * allocating a new resource.
>> */
>> first = region_res->child;
>> - if (!first)
>> - return __alloc_dev_dax_range(dev_dax, dax_region->res.start,
>> - to_alloc);
>
> You probably want to retain the condition above?
>
> Otherwise it removes the ability to create new devices or resizing it , once we
> have zero-ed the last one.
>

A quick unit test that I had stashed here to help explain what I mean:

cd /sys/bus/dax/devices
# dax0.0 is the only dax device in the corresponding dax_region
echo dax0.0 > dax0.0/driver/unbind
echo 0 > dax0.0/size
# Shouldn't fail but returns -ENOSPC despite having
# the full region available
cat $(readlink -f dax0.0/../dax_region/available_size) > dax0.0/size

>> - for (res = first; to_alloc && res; res = res->sibling) {
>> +retry:
>> + rc = -ENOSPC;
>> + for (res = first; res; res = res->sibling) {
>> struct resource *next = res->sibling;
>> - resource_size_t free;
>>
>> /* space at the beginning of the region */
>> - free = 0;
>> - if (res == first && res->start > dax_region->res.start)
>> - free = res->start - dax_region->res.start;
>> - if (free >= to_alloc && dev_size == 0)
>> - return __alloc_dev_dax_range(dev_dax,
>> - dax_region->res.start, to_alloc);
>> -
>> - free = 0;
>> + if (res == first && res->start > dax_region->res.start) {
>> + alloc = min(res->start - dax_region->res.start,
>> + to_alloc);
>> + rc = __alloc_dev_dax_range(dev_dax,
>> + dax_region->res.start, alloc);
>> + break;
>> + }
>> +
>> + alloc = 0;
>> /* space between allocations */
>> if (next && next->start > res->end + 1)
>> - free = next->start - res->end + 1;
>> + alloc = min(next->start - (res->end + 1), to_alloc);
>>
>> /* space at the end of the region */
>> - if (free < to_alloc && !next && res->end < region_res->end)
>> - free = region_res->end - res->end;
>> -
>> - if (free >= to_alloc && strcmp(name, res->name) == 0)
>> - return __adjust_dev_dax_range(dev_dax, res,
>> - resource_size(res) + to_alloc);
>> - else if (free >= to_alloc && dev_size == 0)
>> - return __alloc_dev_dax_range(dev_dax, res->end + 1,
>> - to_alloc);
>> + if (!alloc && !next && res->end < region_res->end)
>> + alloc = min(region_res->end - res->end, to_alloc);
>> +
>> + if (!alloc)
>> + continue;
>> +
>> + if (adjust_ok(dev_dax, res)) {
>> + rc = __adjust_dev_dax_range(dev_dax, res,
>> + resource_size(res) + alloc);
>> + break;
>> + }
>> + rc = __alloc_dev_dax_range(dev_dax, res->end + 1,
>> + alloc);
>
> I am wondering if we should switch to:
>
> if (adjust_ok(...))
> rc = __adjust_dev_dax_range(...);
> else
> rc = __alloc_dev_dax_range(...);
>
> And then a debug print at the end depicting whether and how did we grabbed
> space? Something like:
>
> dev_dbg(&dev_dax->dev, "%s(%d) %d", action, location, rc);
>
> Assuming we set @location to its values when we allocate space at the end,
> beginning or middle; and @action to whether we adjusted up/down or allocated new
> range.
>
> Essentially, something similar to namespaces scan_allocate() just to help
> troubleshoot?
>
> Regards,
> Joao
>