RE: [RFC mm/zswap 1/2] mm/zswap: add the flag can_sleep_mapped

["Song Bao Hua (Barry Song)" <song.bao.hua@hisilicon.com>]

> -----Original Message-----
> From: Vitaly Wool [mailto:vitaly.wool@konsulko.com]
> Sent: Thursday, January 21, 2021 10:17 PM
> To: tiantao (H) <tiantao6@hisilicon.com>
> Cc: Seth Jennings <sjenning@redhat.com>; Dan Streetman <ddstreet@ieee.org>;
> Andrew Morton <akpm@linux-foundation.org>; Song Bao Hua (Barry Song)
> <song.bao.hua@hisilicon.com>; Linux-MM <linux-mm@kvack.org>
> Subject: Re: [RFC mm/zswap 1/2] mm/zswap: add the flag can_sleep_mapped
> 
> On Fri, Dec 25, 2020 at 12:02 PM Tian Tao <tiantao6@hisilicon.com> wrote:
> >
> > add a flag to zpool, named is "can_sleep_mapped", and have it set true
> > for zbud/z3fold, set false for zsmalloc. Then zswap could go the current
> > path if the flag is true; and if it's false, copy data from src to a
> > temporary buffer, then unmap the handle, take the mutex, process the
> > buffer instead of src to avoid sleeping function called from atomic
> > context.
> >
> > Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
> > ---
> >  include/linux/zpool.h |  3 +++
> >  mm/zpool.c            | 13 +++++++++++++
> >  mm/zswap.c            | 50
> +++++++++++++++++++++++++++++++++++++++++++++-----
> >  3 files changed, 61 insertions(+), 5 deletions(-)
> >
> > diff --git a/include/linux/zpool.h b/include/linux/zpool.h
> > index 51bf430..e899701 100644
> > --- a/include/linux/zpool.h
> > +++ b/include/linux/zpool.h
> > @@ -73,6 +73,7 @@ u64 zpool_get_total_size(struct zpool *pool);
> >   * @malloc:    allocate mem from a pool.
> >   * @free:      free mem from a pool.
> >   * @shrink:    shrink the pool.
> > + * @sleep_mapped: whether zpool driver can sleep during map.
> >   * @map:       map a handle.
> >   * @unmap:     unmap a handle.
> >   * @total_size:        get total size of a pool.
> > @@ -100,6 +101,7 @@ struct zpool_driver {
> >         int (*shrink)(void *pool, unsigned int pages,
> >                                 unsigned int *reclaimed);
> >
> > +       bool sleep_mapped;
> >         void *(*map)(void *pool, unsigned long handle,
> >                                 enum zpool_mapmode mm);
> >         void (*unmap)(void *pool, unsigned long handle);
> > @@ -112,5 +114,6 @@ void zpool_register_driver(struct zpool_driver *driver);
> >  int zpool_unregister_driver(struct zpool_driver *driver);
> >
> >  bool zpool_evictable(struct zpool *pool);
> > +bool zpool_can_sleep_mapped(struct zpool *pool);
> >
> >  #endif
> > diff --git a/mm/zpool.c b/mm/zpool.c
> > index 3744a2d..5ed7120 100644
> > --- a/mm/zpool.c
> > +++ b/mm/zpool.c
> > @@ -23,6 +23,7 @@ struct zpool {
> >         void *pool;
> >         const struct zpool_ops *ops;
> >         bool evictable;
> > +       bool can_sleep_mapped;
> >
> >         struct list_head list;
> >  };
> > @@ -183,6 +184,7 @@ struct zpool *zpool_create_pool(const char *type, const
> char *name, gfp_t gfp,
> >         zpool->pool = driver->create(name, gfp, ops, zpool);
> >         zpool->ops = ops;
> >         zpool->evictable = driver->shrink && ops && ops->evict;
> > +       zpool->can_sleep_mapped = driver->sleep_mapped;
> >
> >         if (!zpool->pool) {
> >                 pr_err("couldn't create %s pool\n", type);
> > @@ -393,6 +395,17 @@ bool zpool_evictable(struct zpool *zpool)
> >         return zpool->evictable;
> >  }
> >
> > +/**
> > + * zpool_can_sleep_mapped - Test if zpool can sleep when do mapped.
> > + * @zpool:     The zpool to test
> > + *
> > + * Returns: true if zpool can sleep; false otherwise.
> > + */
> > +bool zpool_can_sleep_mapped(struct zpool *zpool)
> > +{
> > +       return zpool->can_sleep_mapped;
> > +}
> > +
> >  MODULE_LICENSE("GPL");
> >  MODULE_AUTHOR("Dan Streetman <ddstreet@ieee.org>");
> >  MODULE_DESCRIPTION("Common API for compressed memory storage");
> > diff --git a/mm/zswap.c b/mm/zswap.c
> > index 182f6ad..67d4555 100644
> > --- a/mm/zswap.c
> > +++ b/mm/zswap.c
> > @@ -935,13 +935,20 @@ static int zswap_writeback_entry(struct zpool *pool,
> unsigned long handle)
> >         struct scatterlist input, output;
> >         struct crypto_acomp_ctx *acomp_ctx;
> >
> > -       u8 *src;
> > +       u8 *src, *tmp;
> >         unsigned int dlen;
> >         int ret;
> >         struct writeback_control wbc = {
> >                 .sync_mode = WB_SYNC_NONE,
> >         };
> >
> > +       if (!zpool_can_sleep_mapped(pool)) {
> > +
> > +               tmp = kmalloc(entry->length, GFP_ATOMIC);
> 
> This has escaped my attention, but this is obviously a bug. 'entry' is
> not initialized at this point.
> You either have to move memory allocation further down when
> entry->length starts making sense, or allocate PAGE_SIZE.
> 

Since zsmalloc has no evict entry, we have never arrived here.
That's why the testing by both Tiantao and Sebastian didn't
show the problem.


> Best regards,
>    Vitaly
> 
> > +               if (!tmp)
> > +                       return -ENOMEM;
> > +       }
> > +
> >         /* extract swpentry from data */
> >         zhdr = zpool_map_handle(pool, handle, ZPOOL_MM_RO);
> >         swpentry = zhdr->swpentry; /* here */
> > @@ -979,6 +986,14 @@ static int zswap_writeback_entry(struct zpool *pool,
> unsigned long handle)
> >                 dlen = PAGE_SIZE;
> >                 src = (u8 *)zhdr + sizeof(struct zswap_header);
> >
> > +               if (!zpool_can_sleep_mapped(pool)) {
> > +
> > +                       memcpy(tmp, src, entry->length);
> > +                       src = tmp;
> > +
> > +                       zpool_unmap_handle(pool, handle);
> > +               }
> > +
> >                 mutex_lock(acomp_ctx->mutex);
> >                 sg_init_one(&input, src, entry->length);
> >                 sg_init_table(&output, 1);
> > @@ -1033,7 +1048,11 @@ static int zswap_writeback_entry(struct zpool *pool,
> unsigned long handle)
> >         spin_unlock(&tree->lock);
> >
> >  end:
> > -       zpool_unmap_handle(pool, handle);
> > +       if (zpool_can_sleep_mapped(pool))
> > +               zpool_unmap_handle(pool, handle);
> > +       else
> > +               kfree(tmp);
> > +
> >         return ret;
> >  }
> >
> > @@ -1235,7 +1254,7 @@ static int zswap_frontswap_load(unsigned type, pgoff_t
> offset,
> >         struct zswap_entry *entry;
> >         struct scatterlist input, output;
> >         struct crypto_acomp_ctx *acomp_ctx;
> > -       u8 *src, *dst;
> > +       u8 *src, *dst, *tmp;
> >         unsigned int dlen;
> >         int ret;
> >
> > @@ -1256,12 +1275,29 @@ static int zswap_frontswap_load(unsigned type, pgoff_t
> offset,
> >                 goto freeentry;
> >         }
> >
> > +       if (!zpool_can_sleep_mapped(entry->pool->zpool)) {
> > +
> > +               tmp = kmalloc(entry->length, GFP_ATOMIC);
> > +               if (!tmp) {
> > +                       ret = -ENOMEM;
> > +                       goto freeentry;
> > +               }
> > +       }
> > +
> >         /* decompress */
> >         dlen = PAGE_SIZE;
> >         src = zpool_map_handle(entry->pool->zpool, entry->handle,
> ZPOOL_MM_RO);
> >         if (zpool_evictable(entry->pool->zpool))
> >                 src += sizeof(struct zswap_header);
> >
> > +       if (!zpool_can_sleep_mapped(entry->pool->zpool)) {
> > +
> > +               memcpy(tmp, src, entry->length);
> > +               src = tmp;
> > +
> > +               zpool_unmap_handle(entry->pool->zpool, entry->handle);
> > +       }
> > +
> >         acomp_ctx = raw_cpu_ptr(entry->pool->acomp_ctx);
> >         mutex_lock(acomp_ctx->mutex);
> >         sg_init_one(&input, src, entry->length);
> > @@ -1271,7 +1307,11 @@ static int zswap_frontswap_load(unsigned type, pgoff_t
> offset,
> >         ret = crypto_wait_req(crypto_acomp_decompress(acomp_ctx->req),
> &acomp_ctx->wait);
> >         mutex_unlock(acomp_ctx->mutex);
> >
> > -       zpool_unmap_handle(entry->pool->zpool, entry->handle);
> > +       if (zpool_can_sleep_mapped(entry->pool->zpool))
> > +               zpool_unmap_handle(entry->pool->zpool, entry->handle);
> > +       else
> > +               kfree(tmp);
> > +
> >         BUG_ON(ret);
> >
> >  freeentry:
> > @@ -1279,7 +1319,7 @@ static int zswap_frontswap_load(unsigned type, pgoff_t
> offset,
> >         zswap_entry_put(tree, entry);
> >         spin_unlock(&tree->lock);
> >
> > -       return 0;
> > +       return ret;
> >  }
> >
> >  /* frees an entry in zswap */
> > --
> > 2.7.4
> >

Thanks
Barry