Re: [PATCH v1 3/7] xen-blkback: implement LRU mechanism forpersistent grants

From: Konrad Rzeszutek Wilk
Date: Tue Apr 09 2013 - 11:42:20 EST


On Wed, Mar 27, 2013 at 12:10:39PM +0100, Roger Pau Monne wrote:
> This mechanism allows blkback to change the number of grants
> persistently mapped at run time.
>
> The algorithm uses a simple LRU mechanism that removes (if needed) the
> persistent grants that have not been used since the last LRU run, or
> if all grants have been used it removes the first grants in the list
> (that are not in use).
>
> The algorithm allows the user to change the maximum number of
> persistent grants, by changing max_persistent_grants in sysfs.
>
> Signed-off-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>
> Cc: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
> Cc: xen-devel@xxxxxxxxxxxxx
> ---
> Changes since RFC:
> * Offload part of the purge work to a kworker thread.
> * Add documentation about sysfs tunable.
> ---
> Documentation/ABI/stable/sysfs-bus-xen-backend | 7 +
> drivers/block/xen-blkback/blkback.c | 278 +++++++++++++++++++-----
> drivers/block/xen-blkback/common.h | 12 +
> drivers/block/xen-blkback/xenbus.c | 2 +
> 4 files changed, 240 insertions(+), 59 deletions(-)
>
> diff --git a/Documentation/ABI/stable/sysfs-bus-xen-backend b/Documentation/ABI/stable/sysfs-bus-xen-backend
> index e04afe0..7595b38 100644
> --- a/Documentation/ABI/stable/sysfs-bus-xen-backend
> +++ b/Documentation/ABI/stable/sysfs-bus-xen-backend
> @@ -81,3 +81,10 @@ Contact: Roger Pau Monné <roger.pau@xxxxxxxxxx>
> Description:
> Maximum number of free pages to keep in each block
> backend buffer.
> +
> +What: /sys/module/xen_blkback/parameters/max_persistent_grants
> +Date: March 2013
> +KernelVersion: 3.10
> +Contact: Roger Pau Monné <roger.pau@xxxxxxxxxx>
> +Description:
> + Maximum number of grants to map persistently in blkback.

You are missing the other 2 paramters.

> diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
> index 8a1892a..fd1dd38 100644
> --- a/drivers/block/xen-blkback/blkback.c
> +++ b/drivers/block/xen-blkback/blkback.c
> @@ -78,6 +78,44 @@ module_param_named(max_buffer_pages, xen_blkif_max_buffer_pages, int, 0644);
> MODULE_PARM_DESC(max_buffer_pages,
> "Maximum number of free pages to keep in each block backend buffer");
>
> +/*
> + * Maximum number of grants to map persistently in blkback. For maximum
> + * performance this should be the total numbers of grants that can be used
> + * to fill the ring, but since this might become too high, specially with
> + * the use of indirect descriptors, we set it to a value that provides good
> + * performance without using too much memory.
> + *
> + * When the list of persistent grants is full we clean it using a LRU
^- up

> + * algorithm.

Sorry for the confusion. Does this mean that the cap is 352? Or that once
we hit the cap of 352 then the LRU engine gets kicked in?

> + */
> +
> +static int xen_blkif_max_pgrants = 352;
> +module_param_named(max_persistent_grants, xen_blkif_max_pgrants, int, 0644);
> +MODULE_PARM_DESC(max_persistent_grants,
> + "Maximum number of grants to map persistently");
> +
> +/*
> + * The LRU mechanism to clean the lists of persistent grants needs to
> + * be executed periodically. The time interval between consecutive executions
> + * of the purge mechanism is set in ms.
> + */
> +
> +static int xen_blkif_lru_interval = 100;
> +module_param_named(lru_interval, xen_blkif_lru_interval, int, 0);

Is the '0' intentional?

> +MODULE_PARM_DESC(lru_interval,
> +"Execution interval (in ms) of the LRU mechanism to clean the list of persistent grants");
> +
> +/*
> + * When the persistent grants list is full we will remove unused grants
> + * from the list. The percent number of grants to be removed at each LRU
> + * execution.
> + */
> +
> +static int xen_blkif_lru_percent_clean = 5;

OK, so by default if the persistent grants are at 352 pages we start trimming
5% of the list every 100ms. Are we trimming the whole list or just down
to 352 (so max_persistent_grants?).

If the latter do we stop the LRU execution at that point? If so please
enumerate that. Oh and do we stop at 352 or do we do the 5% so we end up actually
at 334?

Would it make sense to cap the engine to stop removing unused grants
down to xen_blkif_max_pgrants?

> +module_param_named(lru_percent_clean, xen_blkif_lru_percent_clean, int, 0);

Ditto here.

> +MODULE_PARM_DESC(lru_percent_clean,
> +"Number of persistent grants to unmap when the list is full");
> +
> /* Run-time switchable: /sys/module/blkback/parameters/ */
> static unsigned int log_stats;
> module_param(log_stats, int, 0644);
> @@ -96,8 +134,8 @@ struct pending_req {
> unsigned short operation;
> int status;
> struct list_head free_list;
> - DECLARE_BITMAP(unmap_seg, BLKIF_MAX_SEGMENTS_PER_REQUEST);
> struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> + struct persistent_gnt *persistent_gnts[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> };
>
> #define BLKBACK_INVALID_HANDLE (~0)
> @@ -119,36 +157,6 @@ struct xen_blkbk {
> static struct xen_blkbk *blkbk;
>
> /*
> - * Maximum number of grant pages that can be mapped in blkback.
> - * BLKIF_MAX_SEGMENTS_PER_REQUEST * RING_SIZE is the maximum number of
> - * pages that blkback will persistently map.
> - * Currently, this is:
> - * RING_SIZE = 32 (for all known ring types)
> - * BLKIF_MAX_SEGMENTS_PER_REQUEST = 11
> - * sizeof(struct persistent_gnt) = 48
> - * So the maximum memory used to store the grants is:
> - * 32 * 11 * 48 = 16896 bytes
> - */
> -static inline unsigned int max_mapped_grant_pages(enum blkif_protocol protocol)
> -{
> - switch (protocol) {
> - case BLKIF_PROTOCOL_NATIVE:
> - return __CONST_RING_SIZE(blkif, PAGE_SIZE) *
> - BLKIF_MAX_SEGMENTS_PER_REQUEST;
> - case BLKIF_PROTOCOL_X86_32:
> - return __CONST_RING_SIZE(blkif_x86_32, PAGE_SIZE) *
> - BLKIF_MAX_SEGMENTS_PER_REQUEST;
> - case BLKIF_PROTOCOL_X86_64:
> - return __CONST_RING_SIZE(blkif_x86_64, PAGE_SIZE) *
> - BLKIF_MAX_SEGMENTS_PER_REQUEST;
> - default:
> - BUG();
> - }
> - return 0;
> -}
> -
> -
> -/*
> * Little helpful macro to figure out the index and virtual address of the
> * pending_pages[..]. For each 'pending_req' we have have up to
> * BLKIF_MAX_SEGMENTS_PER_REQUEST (11) pages. The seg would be from 0 through
> @@ -239,13 +247,19 @@ static void make_response(struct xen_blkif *blkif, u64 id,
> (n) = (&(pos)->node != NULL) ? rb_next(&(pos)->node) : NULL)
>
>
> -static int add_persistent_gnt(struct rb_root *root,
> +static int add_persistent_gnt(struct xen_blkif *blkif,
> struct persistent_gnt *persistent_gnt)
> {
> - struct rb_node **new = &(root->rb_node), *parent = NULL;
> + struct rb_node **new = NULL, *parent = NULL;
> struct persistent_gnt *this;
>
> + if (blkif->persistent_gnt_c >= xen_blkif_max_pgrants) {
> + if (!blkif->vbd.overflow_max_grants)
> + blkif->vbd.overflow_max_grants = 1;
> + return -EBUSY;
> + }
> /* Figure out where to put new node */
> + new = &blkif->persistent_gnts.rb_node;
> while (*new) {
> this = container_of(*new, struct persistent_gnt, node);
>
> @@ -259,19 +273,23 @@ static int add_persistent_gnt(struct rb_root *root,
> return -EINVAL;
> }
> }
> -
> + bitmap_zero(persistent_gnt->flags, PERSISTENT_GNT_FLAGS_SIZE);
> + set_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags);
> /* Add new node and rebalance tree. */
> rb_link_node(&(persistent_gnt->node), parent, new);
> - rb_insert_color(&(persistent_gnt->node), root);
> + rb_insert_color(&(persistent_gnt->node), &blkif->persistent_gnts);
> + blkif->persistent_gnt_c++;
> + atomic_inc(&blkif->persistent_gnt_in_use);
> return 0;
> }
>
> -static struct persistent_gnt *get_persistent_gnt(struct rb_root *root,
> +static struct persistent_gnt *get_persistent_gnt(struct xen_blkif *blkif,
> grant_ref_t gref)
> {
> struct persistent_gnt *data;
> - struct rb_node *node = root->rb_node;
> + struct rb_node *node = NULL;
>
> + node = blkif->persistent_gnts.rb_node;
> while (node) {
> data = container_of(node, struct persistent_gnt, node);
>
> @@ -279,12 +297,25 @@ static struct persistent_gnt *get_persistent_gnt(struct rb_root *root,
> node = node->rb_left;
> else if (gref > data->gnt)
> node = node->rb_right;
> - else
> + else {
> + WARN_ON(test_bit(PERSISTENT_GNT_ACTIVE, data->flags));

Ouch. Is that even safe? Doesn't that mean that somebody is using this and we just
usurped it?
> + set_bit(PERSISTENT_GNT_ACTIVE, data->flags);
> + atomic_inc(&blkif->persistent_gnt_in_use);
> return data;
> + }
> }
> return NULL;
> }
>
> +static void put_persistent_gnt(struct xen_blkif *blkif,
> + struct persistent_gnt *persistent_gnt)
> +{
> + WARN_ON(!test_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags));
> + set_bit(PERSISTENT_GNT_USED, persistent_gnt->flags);
> + clear_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags);
> + atomic_dec(&blkif->persistent_gnt_in_use);
> +}
> +
> static void free_persistent_gnts(struct xen_blkif *blkif, struct rb_root *root,
> unsigned int num)
> {
> @@ -322,6 +353,122 @@ static void free_persistent_gnts(struct xen_blkif *blkif, struct rb_root *root,
> BUG_ON(num != 0);
> }
>

Might need need to say something about locking..

> +static void unmap_purged_grants(struct work_struct *work)
> +{
> + struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> + struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> + struct persistent_gnt *persistent_gnt;
> + int ret, segs_to_unmap = 0;
> + struct xen_blkif *blkif =
> + container_of(work, typeof(*blkif), persistent_purge_work);
> +
> + while(!list_empty(&blkif->persistent_purge_list)) {
> + persistent_gnt = list_first_entry(&blkif->persistent_purge_list,
> + struct persistent_gnt,
> + remove_node);
> + list_del(&persistent_gnt->remove_node);
> +
> + gnttab_set_unmap_op(&unmap[segs_to_unmap],
> + vaddr(persistent_gnt->page),
> + GNTMAP_host_map,
> + persistent_gnt->handle);
> +
> + pages[segs_to_unmap] = persistent_gnt->page;
> +
> + if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
> + ret = gnttab_unmap_refs(unmap, NULL, pages,
> + segs_to_unmap);
> + BUG_ON(ret);
> + put_free_pages(blkif, pages, segs_to_unmap);
> + segs_to_unmap = 0;
> + }
> + kfree(persistent_gnt);
> + }
> + if (segs_to_unmap > 0) {
> + ret = gnttab_unmap_refs(unmap, NULL, pages, segs_to_unmap);
> + BUG_ON(ret);
> + put_free_pages(blkif, pages, segs_to_unmap);
> + }
> +}
> +
> +static void purge_persistent_gnt(struct xen_blkif *blkif)
> +{
> + struct persistent_gnt *persistent_gnt;
> + struct rb_node *n;
> + int num_clean, total;
> + bool scan_used = false;
> + struct rb_root *root;
> +
> + if (blkif->persistent_gnt_c < xen_blkif_max_pgrants ||
> + (blkif->persistent_gnt_c == xen_blkif_max_pgrants &&
> + !blkif->vbd.overflow_max_grants)) {
> + return;
> + }
> +
> + if (work_pending(&blkif->persistent_purge_work)) {
> + pr_alert_ratelimited(DRV_PFX "Scheduled work from previous purge is still pending, cannot purge list\n");
> + return;
> + }
> +
> + num_clean = (xen_blkif_max_pgrants / 100) * xen_blkif_lru_percent_clean;
> + num_clean = blkif->persistent_gnt_c - xen_blkif_max_pgrants + num_clean;
> + num_clean = num_clean > blkif->persistent_gnt_c ?
> + blkif->persistent_gnt_c : num_clean;

I think you can do:
num_clean = min(blkif->persistent_gnt_c, num_clean);
?

> + if (num_clean >
> + (blkif->persistent_gnt_c -
> + atomic_read(&blkif->persistent_gnt_in_use)))
> + return;
> +
> + total = num_clean;
> +
> + pr_debug(DRV_PFX "Going to purge %d persistent grants\n", num_clean);
> +
> + INIT_LIST_HEAD(&blkif->persistent_purge_list);
> + root = &blkif->persistent_gnts;
> +purge_list:
> + foreach_grant_safe(persistent_gnt, n, root, node) {
> + BUG_ON(persistent_gnt->handle ==
> + BLKBACK_INVALID_HANDLE);
> +
> + if (test_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags))
> + continue;
> + if (!scan_used &&
> + (test_bit(PERSISTENT_GNT_USED, persistent_gnt->flags)))
> + continue;
> +
> + rb_erase(&persistent_gnt->node, root);
> + list_add(&persistent_gnt->remove_node,
> + &blkif->persistent_purge_list);
> + if (--num_clean == 0)
> + goto finished;

The check (if num_clean > .. atomic_read) means that by the time we
get here and try to remove some of them, the blkif->persistent_gnt_in_use
might have dramatically changed.

Which would mean we could be walking the list here without collecting
any grants.

Thought on the second pass we would hopefully get the PERSISTENT_GNT_USED
but might also hit that case.

Is that a case we need to worry about? I mean if we hit it, then num_clean
is less than the total. Perhaps at that stage we should decrease
the 'xen_blkif_lru_percent_clean' to a smaller number as we are probably
going to hit the same issue next time around?

And conversely - the next time we run this code and num_clean == total
then we can increase the 'xen_blkif_lru_percent_clean' to its original
value?

That would mean we need somewhere a new value (blkif specific) that
would serve the role of being a local version of 'xen_blkif_lru_percent_clean'
that can be from 0 -> xen_blkif_lru_percent_clean?

Or is this case I am thinking of overblown and you did not hit it when
running stress-tests with this code?



> + }
> + /*
> + * If we get here it means we also need to start cleaning
> + * grants that were used since last purge in order to cope
> + * with the requested num
> + */
> + if (!scan_used) {
> + pr_debug(DRV_PFX "Still missing %d purged frames\n", num_clean);
> + scan_used = true;
> + goto purge_list;
> + }
> +finished:
> + /* Remove the "used" flag from all the persistent grants */
> + foreach_grant_safe(persistent_gnt, n, root, node) {
> + BUG_ON(persistent_gnt->handle ==
> + BLKBACK_INVALID_HANDLE);
> + clear_bit(PERSISTENT_GNT_USED, persistent_gnt->flags);

The 'used' flag does not sound right. I can't place it but I would
think another name is better.

Perhaps 'WALKED' ? 'TOUCHED' ? 'NOT_TOO_FRESH'? 'STALE'? 'WAS_ACTIVE'?

Hm, perhaps 'WAS_ACTIVE'? Or maybe the 'USED' is right.

Either way, I am not a native speaker so lets not get hanged up on this.


> + }
> + blkif->persistent_gnt_c -= (total - num_clean);
> + blkif->vbd.overflow_max_grants = 0;
> +
> + /* We can defer this work */
> + INIT_WORK(&blkif->persistent_purge_work, unmap_purged_grants);
> + schedule_work(&blkif->persistent_purge_work);
> + pr_debug(DRV_PFX "Purged %d/%d\n", (total - num_clean), total);
> + return;
> +}
> +
> /*
> * Retrieve from the 'pending_reqs' a free pending_req structure to be used.
> */
> @@ -453,12 +600,12 @@ irqreturn_t xen_blkif_be_int(int irq, void *dev_id)
> static void print_stats(struct xen_blkif *blkif)
> {
> pr_info("xen-blkback (%s): oo %3llu | rd %4llu | wr %4llu | f %4llu"
> - " | ds %4llu | pg: %4u/%4u\n",
> + " | ds %4llu | pg: %4u/%4d\n",
> current->comm, blkif->st_oo_req,
> blkif->st_rd_req, blkif->st_wr_req,
> blkif->st_f_req, blkif->st_ds_req,
> blkif->persistent_gnt_c,
> - max_mapped_grant_pages(blkif->blk_protocol));
> + xen_blkif_max_pgrants);
> blkif->st_print = jiffies + msecs_to_jiffies(10 * 1000);
> blkif->st_rd_req = 0;
> blkif->st_wr_req = 0;
> @@ -470,6 +617,7 @@ int xen_blkif_schedule(void *arg)
> {
> struct xen_blkif *blkif = arg;
> struct xen_vbd *vbd = &blkif->vbd;
> + unsigned long timeout;
>
> xen_blkif_get(blkif);
>
> @@ -479,13 +627,21 @@ int xen_blkif_schedule(void *arg)
> if (unlikely(vbd->size != vbd_sz(vbd)))
> xen_vbd_resize(blkif);
>
> - wait_event_interruptible(
> + timeout = msecs_to_jiffies(xen_blkif_lru_interval);
> +
> + timeout = wait_event_interruptible_timeout(
> blkif->wq,
> - blkif->waiting_reqs || kthread_should_stop());
> - wait_event_interruptible(
> + blkif->waiting_reqs || kthread_should_stop(),
> + timeout);
> + if (timeout == 0)
> + goto purge_gnt_list;

So if the system admin decided that the value of zero in xen_blkif_lru_interval
was perfect, won't we timeout immediately and keep on going to the goto statement,
and then looping around?

I think we need some sanity checking that the xen_blkif_lru_interval and
perhaps the other parameters are sane. And sadly we cannot do that
in the init function as the system admin might do this:

for fun in `ls -1 /sys/modules/xen_blkback/parameters`
do
echo 0 > /sys/modules/xen_blkback/parameters/$a
done

and reset everything to zero. Yikes.


> + timeout = wait_event_interruptible_timeout(

if (blkif->vbd.feature_gnt_persistent) {
timeout = msecs_to_jiffiers(..)
timeout = wait_event_interruptple_timeout
> blkbk->pending_free_wq,
> !list_empty(&blkbk->pending_free) ||
> - kthread_should_stop());
> + kthread_should_stop(),
> + timeout);
> + if (timeout == 0)
> + goto purge_gnt_list;
>
> blkif->waiting_reqs = 0;
> smp_mb(); /* clear flag *before* checking for work */
> @@ -493,6 +649,14 @@ int xen_blkif_schedule(void *arg)
> if (do_block_io_op(blkif))
> blkif->waiting_reqs = 1;
>
> +purge_gnt_list:
> + if (blkif->vbd.feature_gnt_persistent &&
> + time_after(jiffies, blkif->next_lru)) {
> + purge_persistent_gnt(blkif);
> + blkif->next_lru = jiffies +
> + msecs_to_jiffies(xen_blkif_lru_interval);

You can make this more than 80 characters.

> + }
> +
> shrink_free_pagepool(blkif, xen_blkif_max_buffer_pages);
>
> if (log_stats && time_after(jiffies, blkif->st_print))
> @@ -504,7 +668,7 @@ int xen_blkif_schedule(void *arg)
> /* Free all persistent grant pages */
> if (!RB_EMPTY_ROOT(&blkif->persistent_gnts))
> free_persistent_gnts(blkif, &blkif->persistent_gnts,
> - blkif->persistent_gnt_c);
> + blkif->persistent_gnt_c);
>
> BUG_ON(!RB_EMPTY_ROOT(&blkif->persistent_gnts));
> blkif->persistent_gnt_c = 0;
> @@ -536,8 +700,10 @@ static void xen_blkbk_unmap(struct pending_req *req)
> int ret;
>
> for (i = 0; i < req->nr_pages; i++) {
> - if (!test_bit(i, req->unmap_seg))
> + if (req->persistent_gnts[i] != NULL) {
> + put_persistent_gnt(blkif, req->persistent_gnts[i]);
> continue;
> + }

Aha, the req->unmap_seg bitmap is not used here anymore! You might want to
mention that in the commit description saying that you are also removing
this artifact as you don't need it anymore with the superior cleaning
method employed here.



> handle = pending_handle(req, i);
> pages[invcount] = req->pages[i];
> if (handle == BLKBACK_INVALID_HANDLE)
> @@ -559,8 +725,8 @@ static int xen_blkbk_map(struct blkif_request *req,
> struct page *pages[])
> {
> struct gnttab_map_grant_ref map[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> - struct persistent_gnt *persistent_gnts[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> struct page *pages_to_gnt[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> + struct persistent_gnt **persistent_gnts = pending_req->persistent_gnts;
> struct persistent_gnt *persistent_gnt = NULL;
> struct xen_blkif *blkif = pending_req->blkif;
> phys_addr_t addr = 0;
> @@ -572,9 +738,6 @@ static int xen_blkbk_map(struct blkif_request *req,
>
> use_persistent_gnts = (blkif->vbd.feature_gnt_persistent);
>
> - BUG_ON(blkif->persistent_gnt_c >
> - max_mapped_grant_pages(pending_req->blkif->blk_protocol));
> -
> /*
> * Fill out preq.nr_sects with proper amount of sectors, and setup
> * assign map[..] with the PFN of the page in our domain with the
> @@ -585,7 +748,7 @@ static int xen_blkbk_map(struct blkif_request *req,
>
> if (use_persistent_gnts)
> persistent_gnt = get_persistent_gnt(
> - &blkif->persistent_gnts,
> + blkif,
> req->u.rw.seg[i].gref);
>
> if (persistent_gnt) {
> @@ -621,7 +784,6 @@ static int xen_blkbk_map(struct blkif_request *req,
> * so that when we access vaddr(pending_req,i) it has the contents of
> * the page from the other domain.
> */
> - bitmap_zero(pending_req->unmap_seg, BLKIF_MAX_SEGMENTS_PER_REQUEST);
> for (i = 0, j = 0; i < nseg; i++) {
> if (!persistent_gnts[i]) {
> /* This is a newly mapped grant */
> @@ -639,11 +801,10 @@ static int xen_blkbk_map(struct blkif_request *req,
> if (persistent_gnts[i])
> goto next;
> if (use_persistent_gnts &&
> - blkif->persistent_gnt_c <
> - max_mapped_grant_pages(blkif->blk_protocol)) {
> + blkif->persistent_gnt_c < xen_blkif_max_pgrants) {
> /*
> * We are using persistent grants, the grant is
> - * not mapped but we have room for it
> + * not mapped but we might have room for it.
> */
> persistent_gnt = kmalloc(sizeof(struct persistent_gnt),
> GFP_KERNEL);
> @@ -659,16 +820,16 @@ static int xen_blkbk_map(struct blkif_request *req,
> persistent_gnt->gnt = map[j].ref;
> persistent_gnt->handle = map[j].handle;
> persistent_gnt->page = pages[i];
> - if (add_persistent_gnt(&blkif->persistent_gnts,
> + if (add_persistent_gnt(blkif,
> persistent_gnt)) {
> kfree(persistent_gnt);
> j++;
> goto next;
> }
> - blkif->persistent_gnt_c++;
> + persistent_gnts[i] = persistent_gnt;
> pr_debug(DRV_PFX " grant %u added to the tree of persistent grants, using %u/%u\n",
> persistent_gnt->gnt, blkif->persistent_gnt_c,
> - max_mapped_grant_pages(blkif->blk_protocol));
> + xen_blkif_max_pgrants);
> j++;
> goto next;
> }
> @@ -677,7 +838,6 @@ static int xen_blkbk_map(struct blkif_request *req,
> pr_debug(DRV_PFX " domain %u, device %#x is using maximum number of persistent grants\n",
> blkif->domid, blkif->vbd.handle);
> }
> - bitmap_set(pending_req->unmap_seg, i, 1);
> j++;
> next:
> seg[i].offset = (req->u.rw.seg[i].first_sect << 9);
> diff --git a/drivers/block/xen-blkback/common.h b/drivers/block/xen-blkback/common.h
> index 6c73c38..f43782c 100644
> --- a/drivers/block/xen-blkback/common.h
> +++ b/drivers/block/xen-blkback/common.h
> @@ -182,12 +182,17 @@ struct xen_vbd {
>
> struct backend_info;
>
> +#define PERSISTENT_GNT_FLAGS_SIZE 2
> +#define PERSISTENT_GNT_ACTIVE 0
> +#define PERSISTENT_GNT_USED 1

These might need a bit of explanation. Would it be possible to include that
here?

>
> struct persistent_gnt {
> struct page *page;
> grant_ref_t gnt;
> grant_handle_t handle;
> + DECLARE_BITMAP(flags, PERSISTENT_GNT_FLAGS_SIZE);
> struct rb_node node;
> + struct list_head remove_node;
> };
>
> struct xen_blkif {
> @@ -219,6 +224,12 @@ struct xen_blkif {
> /* tree to store persistent grants */
> struct rb_root persistent_gnts;
> unsigned int persistent_gnt_c;
> + atomic_t persistent_gnt_in_use;
> + unsigned long next_lru;
> +
> + /* used by the kworker that offload work from the persistent purge */
> + struct list_head persistent_purge_list;
> + struct work_struct persistent_purge_work;
>
> /* buffer of free pages to map grant refs */
> spinlock_t free_pages_lock;
> @@ -262,6 +273,7 @@ int xen_blkif_xenbus_init(void);
>
> irqreturn_t xen_blkif_be_int(int irq, void *dev_id);
> int xen_blkif_schedule(void *arg);
> +int xen_blkif_purge_persistent(void *arg);
>
> int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
> struct backend_info *be, int state);
> diff --git a/drivers/block/xen-blkback/xenbus.c b/drivers/block/xen-blkback/xenbus.c
> index 24f7f6d..e0fd92a 100644
> --- a/drivers/block/xen-blkback/xenbus.c
> +++ b/drivers/block/xen-blkback/xenbus.c
> @@ -98,6 +98,7 @@ static void xen_update_blkif_status(struct xen_blkif *blkif)
> err = PTR_ERR(blkif->xenblkd);
> blkif->xenblkd = NULL;
> xenbus_dev_error(blkif->be->dev, err, "start xenblkd");
> + return;
> }
> }
>
> @@ -121,6 +122,7 @@ static struct xen_blkif *xen_blkif_alloc(domid_t domid)
> spin_lock_init(&blkif->free_pages_lock);
> INIT_LIST_HEAD(&blkif->free_pages);
> blkif->free_pages_num = 0;
> + atomic_set(&blkif->persistent_gnt_in_use, 0);
>
> return blkif;
> }
> --
> 1.7.7.5 (Apple Git-26)
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/