mm: swap: don't delay swap free for fast swap devices
There are couple of issues with swapcache usage when ZRAM is used
as swap device.
1) Kernel does a swap readahead which can be around 6 to 8 pages
depending on total ram, which is not required for zram since
accesses are fast.
2) Kernel delays the freeing up of swapcache expecting a later hit,
which again is useless in the case of zram.
3) This is not related to swapcache, but zram usage itself.
As mentioned in (2) kernel delays freeing of swapcache, but along with
that it delays zram compressed page free also. i.e. there can be 2 copies,
though one is compressed.
This patch addresses these issues using two new flags
QUEUE_FLAG_FAST and SWP_FAST, to indicate that accesses to the device
will be fast and cheap, and instructs the swap layer to free up
swap space agressively, and not to do read ahead.
Change-Id: I5d2d5176a5f9420300bb2f843f6ecbdb25ea80e4
Signed-off-by: Vinayak Menon <vinmenon@codeaurora.org>
diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index c9914d65..3868665 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -1271,6 +1271,7 @@
zram->disk->private_data = zram;
snprintf(zram->disk->disk_name, 16, "zram%d", device_id);
+ __set_bit(QUEUE_FLAG_FAST, &zram->disk->queue->queue_flags);
/* Actual capacity set using syfs (/sys/block/zram<id>/disksize */
set_capacity(zram->disk, 0);
/* zram devices sort of resembles non-rotational disks */
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index fb910c6..0693c3e 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -508,6 +508,7 @@
#define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
#define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
#define QUEUE_FLAG_DAX 26 /* device supports DAX */
+#define QUEUE_FLAG_FAST 27 /* fast block device (e.g. ram based) */
#define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
(1 << QUEUE_FLAG_STACKABLE) | \
@@ -598,6 +599,7 @@
#define blk_queue_secure_erase(q) \
(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
#define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
+#define blk_queue_fast(q) test_bit(QUEUE_FLAG_FAST, &(q)->queue_flags)
#define blk_noretry_request(rq) \
((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
diff --git a/include/linux/swap.h b/include/linux/swap.h
index 55ff559..d0af005 100644
--- a/include/linux/swap.h
+++ b/include/linux/swap.h
@@ -151,8 +151,9 @@
SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
SWP_STABLE_WRITES = (1 << 10), /* no overwrite PG_writeback pages */
+ SWP_FAST = (1 << 11), /* blkdev access is fast and cheap */
/* add others here before... */
- SWP_SCANNING = (1 << 11), /* refcount in scan_swap_map */
+ SWP_SCANNING = (1 << 12), /* refcount in scan_swap_map */
};
#define SWAP_CLUSTER_MAX 32UL
@@ -389,10 +390,18 @@
/* linux/mm/swapfile.c */
extern atomic_long_t nr_swap_pages;
extern long total_swap_pages;
+extern bool is_swap_fast(swp_entry_t entry);
/* Swap 50% full? Release swapcache more aggressively.. */
-static inline bool vm_swap_full(void)
+static inline bool vm_swap_full(struct swap_info_struct *si)
{
+ /*
+ * If the swap device is fast, return true
+ * not to delay swap free.
+ */
+ if (si->flags & SWP_FAST)
+ return true;
+
return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
}
@@ -428,7 +437,7 @@
#define get_nr_swap_pages() 0L
#define total_swap_pages 0L
#define total_swapcache_pages() 0UL
-#define vm_swap_full() 0
+#define vm_swap_full(si) 0
#define si_swapinfo(val) \
do { (val)->freeswap = (val)->totalswap = 0; } while (0)
@@ -579,7 +588,7 @@
static inline bool mem_cgroup_swap_full(struct page *page)
{
- return vm_swap_full();
+ return vm_swap_full(page_swap_info(page));
}
#endif
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index fdc790a..3b38b73 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -5996,7 +5996,7 @@
VM_BUG_ON_PAGE(!PageLocked(page), page);
- if (vm_swap_full())
+ if (vm_swap_full(page_swap_info(page)))
return true;
if (!do_swap_account || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
return false;
diff --git a/mm/swap_state.c b/mm/swap_state.c
index 35d7e0e..8d8b850 100644
--- a/mm/swap_state.c
+++ b/mm/swap_state.c
@@ -475,7 +475,8 @@
unsigned long entry_offset = swp_offset(entry);
unsigned long offset = entry_offset;
unsigned long start_offset, end_offset;
- unsigned long mask;
+ unsigned long mask = is_swap_fast(entry) ? 0 :
+ (1UL << page_cluster) - 1;
struct blk_plug plug;
mask = swapin_nr_pages(offset) - 1;
diff --git a/mm/swapfile.c b/mm/swapfile.c
index d76b2a1..daf1480 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -97,6 +97,26 @@
return ent & ~SWAP_HAS_CACHE; /* may include SWAP_HAS_CONT flag */
}
+bool is_swap_fast(swp_entry_t entry)
+{
+ struct swap_info_struct *p;
+ unsigned long type;
+
+ if (non_swap_entry(entry))
+ return false;
+
+ type = swp_type(entry);
+ if (type >= nr_swapfiles)
+ return false;
+
+ p = swap_info[type];
+
+ if (p->flags & SWP_FAST)
+ return true;
+
+ return false;
+}
+
/* returns 1 if swap entry is freed */
static int
__try_to_reclaim_swap(struct swap_info_struct *si, unsigned long offset)
@@ -573,7 +593,7 @@
scan_base = offset = si->lowest_bit;
/* reuse swap entry of cache-only swap if not busy. */
- if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
+ if (vm_swap_full(si) && si->swap_map[offset] == SWAP_HAS_CACHE) {
int swap_was_freed;
spin_unlock(&si->lock);
swap_was_freed = __try_to_reclaim_swap(si, offset);
@@ -613,7 +633,8 @@
spin_lock(&si->lock);
goto checks;
}
- if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
+ if (vm_swap_full(si) &&
+ si->swap_map[offset] == SWAP_HAS_CACHE) {
spin_lock(&si->lock);
goto checks;
}
@@ -628,7 +649,8 @@
spin_lock(&si->lock);
goto checks;
}
- if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
+ if (vm_swap_full(si) &&
+ si->swap_map[offset] == SWAP_HAS_CACHE) {
spin_lock(&si->lock);
goto checks;
}
@@ -2537,6 +2559,9 @@
pr_err("swapon: discard_swap(%p): %d\n",
p, err);
}
+
+ if (blk_queue_fast(bdev_get_queue(p->bdev)))
+ p->flags |= SWP_FAST;
}
mutex_lock(&swapon_mutex);