LU-13212 osc: fall back to vmalloc for large RPCs

For large RPC sizes (16MB+) the page array (4096 brw_page) can
become very large (128KB+ with fscrypt) and should fall back to
vmalloc() if kmalloc() fails due to memory fragmentation.

The mdc/mdt allocations are currently limited to 1MB for readdir
RPCs, but it doesn't hurt to prepare them for larger RPCs from
clients in the future if this limit is increased.

Fixes: 51b32ac2b9b8 ("LU-7990 rpc: increase bulk size")
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I56805f5701d6850412664ce0681a1456b9405580
Reviewed-on: https://review.whamcloud.com/43281
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Bobi Jam <bobijam@hotmail.com>
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Reviewed-by: James Simmons <jsimmons@infradead.org>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/mdc/mdc_request.c b/lustre/mdc/mdc_request.c
index 3ac4b82..0ce7833 100644 (file)
--- a/lustre/mdc/mdc_request.c
+++ b/lustre/mdc/mdc_request.c
@@ -1307,7 +1307,7 @@ static int mdc_read_page_remote(void *data, struct page *page0)
        fid = &op_data->op_fid1;
        LASSERT(inode != NULL);
 
-       OBD_ALLOC_PTR_ARRAY(page_pool, max_pages);
+       OBD_ALLOC_PTR_ARRAY_LARGE(page_pool, max_pages);
        if (page_pool != NULL) {
                page_pool[0] = page0;
        } else {
@@ -1376,7 +1376,7 @@ static int mdc_read_page_remote(void *data, struct page *page0)
        }
 
        if (page_pool != &page0)
-               OBD_FREE_PTR_ARRAY(page_pool, max_pages);
+               OBD_FREE_PTR_ARRAY_LARGE(page_pool, max_pages);
 
        RETURN(rc);
 }

diff --git a/lustre/mdt/mdt_handler.c b/lustre/mdt/mdt_handler.c
index e953635..e8d9582 100644 (file)
--- a/lustre/mdt/mdt_handler.c
+++ b/lustre/mdt/mdt_handler.c
@@ -2535,7 +2535,7 @@ static int mdt_readpage(struct tgt_session_info *tsi)
                                exp_max_brw_size(tsi->tsi_exp));
        rdpg->rp_npages = (rdpg->rp_count + PAGE_SIZE - 1) >>
                          PAGE_SHIFT;
-       OBD_ALLOC_PTR_ARRAY(rdpg->rp_pages, rdpg->rp_npages);
+       OBD_ALLOC_PTR_ARRAY_LARGE(rdpg->rp_pages, rdpg->rp_npages);
        if (rdpg->rp_pages == NULL)
                RETURN(-ENOMEM);
 
@@ -2559,7 +2559,7 @@ free_rdpg:
        for (i = 0; i < rdpg->rp_npages; i++)
                if (rdpg->rp_pages[i] != NULL)
                        __free_page(rdpg->rp_pages[i]);
-       OBD_FREE_PTR_ARRAY(rdpg->rp_pages, rdpg->rp_npages);
+       OBD_FREE_PTR_ARRAY_LARGE(rdpg->rp_pages, rdpg->rp_npages);
 
        if (OBD_FAIL_CHECK(OBD_FAIL_MDS_SENDPAGE))
                RETURN(0);

diff --git a/lustre/mgc/mgc_request.c b/lustre/mgc/mgc_request.c
index ecce505..ba91824 100644 (file)
--- a/lustre/mgc/mgc_request.c
+++ b/lustre/mgc/mgc_request.c
@@ -1654,7 +1654,7 @@ static int mgc_process_recover_nodemap_log(struct obd_device *obd,
        if (cfg->cfg_last_idx == 0 || cld_is_nodemap(cld))
                nrpages = CONFIG_READ_NRPAGES_INIT;
 
-       OBD_ALLOC_PTR_ARRAY(pages, nrpages);
+       OBD_ALLOC_PTR_ARRAY_LARGE(pages, nrpages);
        if (pages == NULL)
                GOTO(out, rc = -ENOMEM);
 
@@ -1824,7 +1824,7 @@ out:
                                break;
                        __free_page(pages[i]);
                }
-               OBD_FREE_PTR_ARRAY(pages, nrpages);
+               OBD_FREE_PTR_ARRAY_LARGE(pages, nrpages);
        }
        return rc;
 }

diff --git a/lustre/mgs/mgs_nids.c b/lustre/mgs/mgs_nids.c
index 94149c5..42b7309 100644 (file)
--- a/lustre/mgs/mgs_nids.c
+++ b/lustre/mgs/mgs_nids.c
@@ -636,7 +636,7 @@ int mgs_get_ir_logs(struct ptlrpc_request *req)
        CDEBUG(D_MGS, "Reading IR log %s bufsize %ld.\n",
               body->mcb_name, bufsize);
 
-       OBD_ALLOC_PTR_ARRAY(pages, nrpages);
+       OBD_ALLOC_PTR_ARRAY_LARGE(pages, nrpages);
        if (!pages)
                GOTO(out, rc = -ENOMEM);
 
@@ -682,7 +682,7 @@ out:
                        __free_page(pages[i]);
                }
 
-               OBD_FREE_PTR_ARRAY(pages, nrpages);
+               OBD_FREE_PTR_ARRAY_LARGE(pages, nrpages);
        }
 
        if (fsdb)

diff --git a/lustre/obdecho/echo_client.c b/lustre/obdecho/echo_client.c
index d35fbb9..d6cc549 100644 (file)
--- a/lustre/obdecho/echo_client.c
+++ b/lustre/obdecho/echo_client.c
@@ -2576,13 +2576,13 @@ static int echo_client_kbrw(struct echo_device *ed, int rw, struct obdo *oa,
        if (rw == OBD_BRW_WRITE)
                brw_flags = OBD_BRW_ASYNC;
 
-       OBD_ALLOC_PTR_ARRAY(pga, npages);
+       OBD_ALLOC_PTR_ARRAY_LARGE(pga, npages);
        if (!pga)
                RETURN(-ENOMEM);
 
-       OBD_ALLOC_PTR_ARRAY(pages, npages);
+       OBD_ALLOC_PTR_ARRAY_LARGE(pages, npages);
        if (!pages) {
-               OBD_FREE_PTR_ARRAY(pga, npages);
+               OBD_FREE_PTR_ARRAY_LARGE(pga, npages);
                RETURN(-ENOMEM);
        }
 
@@ -2634,8 +2634,8 @@ static int echo_client_kbrw(struct echo_device *ed, int rw, struct obdo *oa,
                }
                __free_page(pgp->pg);
        }
-       OBD_FREE_PTR_ARRAY(pga, npages);
-       OBD_FREE_PTR_ARRAY(pages, npages);
+       OBD_FREE_PTR_ARRAY_LARGE(pga, npages);
+       OBD_FREE_PTR_ARRAY_LARGE(pages, npages);
        RETURN(rc);
 }
 

diff --git a/lustre/osc/osc_request.c b/lustre/osc/osc_request.c
index 6cd0a2c..97062a2 100644 (file)
--- a/lustre/osc/osc_request.c
+++ b/lustre/osc/osc_request.c
@@ -2301,7 +2301,7 @@ static void sort_brw_pages(struct brw_page **array, int num)
 static void osc_release_ppga(struct brw_page **ppga, size_t count)
 {
        LASSERT(ppga != NULL);
-       OBD_FREE_PTR_ARRAY(ppga, count);
+       OBD_FREE_PTR_ARRAY_LARGE(ppga, count);
 }
 
 static int brw_interpret(const struct lu_env *env,
@@ -2501,7 +2501,7 @@ int osc_build_rpc(const struct lu_env *env, struct client_obd *cli,
        if (mem_tight)
                mpflag = memalloc_noreclaim_save();
 
-       OBD_ALLOC_PTR_ARRAY(pga, page_count);
+       OBD_ALLOC_PTR_ARRAY_LARGE(pga, page_count);
        if (pga == NULL)
                GOTO(out, rc = -ENOMEM);