staging/lustre/llite: Don't set page writeback on non-dirty page

New writeback changes in 4.2-RC1 have exposed that we incorrectly
set page_writeback on a page that is being written synchronously,
which aside from this new crash (dereference of NULL inode->i_wb
from set_page_writeback) likely threw off some related page
statistics in the past.

 BUG: unable to handle kernel NULL pointer dereference at 0000000000000138
 IP: [<ffffffff8140d90a>] __percpu_counter_add+0x1a/0x80
 PGD 0
 Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
 Modules linked in: osc(C) lmv(C) fld(C) mgc(C) lustre(C) mdc(C) fid(C) lov(C) ksocklnd(C) ptlrpc(C) obdclass(C) lnet(C) libcfs(C) loop sha512_generic crc32 rpcsec_gss_krb5 microcode joydev i2c_piix4 acpi_cpufreq pcspkr nfsd syscopyarea sysfillrect sysimgblt drm_kms_helper ttm drm serio_raw virtio_blk [last unloaded: libcfs]
 CPU: 0 PID: 13328 Comm: cvs Tainted: G         C      4.2.0-rc1-vm-nfs+ #30
 Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
 task: ffff8800cc98a400 ti: ffff8801157e8000 task.ti: ffff8801157e8000
 RIP: 0010:[<ffffffff8140d90a>]  [<ffffffff8140d90a>] __percpu_counter_add+0x1a/0x80
 RSP: 0018:ffff8801157eb698  EFLAGS: 00010086
 RAX: 0000000000000003 RBX: ffffea0002b91cc0 RCX: 000000000000001a
 RDX: 0000000000000020 RSI: 0000000000000001 RDI: 00000000000000e8
 RBP: ffff8801157eb6b8 R08: 0000000000000001 R09: 0000000000000000
 R10: 0000000000000000 R11: 0000000000000000 R12: 00000000000000e8
 R13: 0000000000000001 R14: ffff8800673587a8 R15: ffff8800673589b0
 FS:  00007f6718b89800(0000) GS:ffff88011f400000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 0000000000000138 CR3: 000000009d51c000 CR4: 00000000000007f0
 Stack:
  ffffffff811919e2 ffffea0002b91cc0 ffff880067358998 ffff880119419800
  ffff8801157eb718 ffffffff81191a58 ffff8801157eb788 0000000000000282
  ffff8800ce5ce920 0000000000000000 ffff8800a525af80 ffff880053f68f10
 Call Trace:
  [<ffffffff811919e2>] ? __test_set_page_writeback+0x72/0x240
  [<ffffffff81191a58>] __test_set_page_writeback+0xe8/0x240
  [<ffffffffa04e7a13>] vvp_page_prep_write+0x33/0xb0 [lustre]
  [<ffffffffa028ad87>] cl_page_invoke+0x57/0x90 [obdclass]
  [<ffffffffa028cc8d>] cl_page_prep+0x2d/0x180 [obdclass]
  [<ffffffffa0545d84>] osc_io_submit+0x134/0x4a0 [osc]
  [<ffffffffa02933a3>] cl_io_submit_rw+0x53/0xb0 [obdclass]
  [<ffffffffa0483635>] lov_io_submit+0x3a5/0x570 [lov]
  [<ffffffff810cb7bb>] ? lockdep_init_map+0x5b/0x6d0
  [<ffffffffa02933a3>] cl_io_submit_rw+0x53/0xb0 [obdclass]
  [<ffffffffa029428d>] cl_io_submit_sync+0xed/0x1c0 [obdclass]
  [<ffffffffa04e881d>] vvp_page_sync_io.isra.15+0x4d/0x100 [lustre]
  [<ffffffffa028dd2f>] ? cl_page_clip+0xff/0x130 [obdclass]
  [<ffffffffa04e9f38>] vvp_io_commit_write+0x448/0x500 [lustre]
  [<ffffffffa02939fa>] cl_io_commit_write+0x9a/0x130 [obdclass]
  [<ffffffffa04cb222>] ll_commit_write+0xc2/0x230 [lustre]
  [<ffffffffa04dbdaa>] ll_write_end+0x2a/0x50 [lustre]
  [<ffffffff811851ba>] generic_perform_write+0xfa/0x1b0
  [<ffffffff8121fd2e>] ? dentry_needs_remove_privs.part.16+0x1e/0x30
  [<ffffffff811877d0>] __generic_file_write_iter+0x190/0x1f0
  [<ffffffff8118791a>] generic_file_write_iter+0xea/0x1e0
  [<ffffffffa04e8dd0>] vvp_io_write_start+0xa0/0x1e0 [lustre]
  [<ffffffffa0292469>] cl_io_start+0x49/0x80 [obdclass]
  [<ffffffffa0294803>] cl_io_loop+0x73/0xd0 [obdclass]
  [<ffffffffa04b4ebf>] ll_file_io_generic+0x45f/0x4b0 [lustre]
  [<ffffffffa04b504c>] ll_file_write_iter+0x6c/0xc0 [lustre]
  [<ffffffff8120330a>] __vfs_write+0xaa/0xe0
  [<ffffffff81203969>] vfs_write+0xa9/0x190
  [<ffffffff812046c9>] SyS_write+0x49/0xa0
  [<ffffffff81796572>] entry_SYSCALL_64_fastpath+0x16/0x7a
 Code: 5b 41 5c 41 5d 41 5e 5d c3 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 55 41 54 53 49 89 fc 49 89 f5 48 83 ec 08 65 ff 05 8e d1 bf 7e <48> 8b 47 50 48 63 ca 65 8b 18 48 63 db 48 01 f3 48 39 cb 7d 0a

Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Signed-off-by: Oleg Drokin <oleg.drokin@intel.com>
Reviewed-on: http://review.whamcloud.com/15610
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-6854
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/staging/lustre/lustre/llite/vvp_page.c b/drivers/staging/lustre/lustre/llite/vvp_page.c
index 954ed08c6af2f740cb879a85468d872d4e9fd110..a3cf5ad20c6036b215092ee3c91bc5a7d2145770 100644 (file)
--- a/drivers/staging/lustre/lustre/llite/vvp_page.c
+++ b/drivers/staging/lustre/lustre/llite/vvp_page.c
@@ -227,11 +227,16 @@ static int vvp_page_prep_write(const struct lu_env *env,
                               struct cl_io *unused)
 {
        struct page *vmpage = cl2vm_page(slice);
+       struct cl_page *pg = slice->cpl_page;
 
        LASSERT(PageLocked(vmpage));
        LASSERT(!PageDirty(vmpage));
 
-       set_page_writeback(vmpage);
+       /* ll_writepage path is not a sync write, so need to set page writeback
+        * flag */
+       if (!pg->cp_sync_io)
+               set_page_writeback(vmpage);
+
        vvp_write_pending(cl2ccc(slice->cpl_obj), cl2ccc_page(slice));
 
        return 0;
@@ -298,9 +303,6 @@ static void vvp_page_completion_write(const struct lu_env *env,
        struct cl_page  *pg     = slice->cpl_page;
        struct page      *vmpage = cp->cpg_page;
 
-       LASSERT(ergo(pg->cp_sync_io != NULL, PageLocked(vmpage)));
-       LASSERT(PageWriteback(vmpage));
-
        CL_PAGE_HEADER(D_PAGE, env, pg, "completing WRITE with %d\n", ioret);
 
        /*
@@ -316,14 +318,19 @@ static void vvp_page_completion_write(const struct lu_env *env,
        cp->cpg_write_queued = 0;
        vvp_write_complete(cl2ccc(slice->cpl_obj), cp);
 
-       /*
-        * Only mark the page error only when it's an async write because
-        * applications won't wait for IO to finish.
-        */
-       if (pg->cp_sync_io == NULL)
+       if (pg->cp_sync_io != NULL) {
+               LASSERT(PageLocked(vmpage));
+               LASSERT(!PageWriteback(vmpage));
+       } else {
+               LASSERT(PageWriteback(vmpage));
+               /*
+                * Only mark the page error only when it's an async write
+                * because applications won't wait for IO to finish.
+                */
                vvp_vmpage_error(ccc_object_inode(pg->cp_obj), vmpage, ioret);
 
-       end_page_writeback(vmpage);
+               end_page_writeback(vmpage);
+       }
 }
 
 /**

diff --git a/drivers/staging/lustre/lustre/obdclass/cl_page.c b/drivers/staging/lustre/lustre/obdclass/cl_page.c
index 0f3417159fc10798c14cc1fc16e4abc3f91e04d8..d5fb81f84cd4effa1d625206ec39255c636dde04 100644 (file)
--- a/drivers/staging/lustre/lustre/obdclass/cl_page.c
+++ b/drivers/staging/lustre/lustre/obdclass/cl_page.c
@@ -1192,9 +1192,6 @@ int cl_page_prep(const struct lu_env *env, struct cl_io *io,
        if (result == 0)
                cl_page_io_start(env, pg, crt);
 
-       KLASSERT(ergo(crt == CRT_WRITE && pg->cp_type == CPT_CACHEABLE,
-                     equi(result == 0,
-                          PageWriteback(cl_page_vmpage(env, pg)))));
        CL_PAGE_HEADER(D_TRACE, env, pg, "%d %d\n", crt, result);
        return result;
 }