mmc: core: add timeout for context_info->is_done_rcv

  After dequeue a rq from blk and a accidental rq missing makes
bq->queue fails to fetch blk_rq again since !rq w/o been done.
Whatsoever was that, MUST ensure a schedule mechanisam to break
the dead loop, and issue the missing one again. We observe this
problem, all processes calling aio page writeback or commiting
journal were confined in io_schedule waiting for previous blk rq
to be completed, and cause os stucked.

page:c1337420 count:3 mapcount:0 mapping:dc8fdc34 index:0x502
page flags: 0x4000286c(referenced|uptodate|lru|active|private|writeback)
sleep_on_page: inode: dc8fdb58, 119091
CPU: 3 PID: 199 Comm: abc Not tainted 3.10.0 #298
[<c0013e44>] (unwind_backtrace+0x0/0xe0) from [<c001175c>] (show_stack+0x10/0x14)
[<c001175c>] (show_stack+0x10/0x14) from [<c00cadb8>] (sleep_on_page+0x6c/0x8c)
[<c00cadb8>] (sleep_on_page+0x6c/0x8c) from [<c0769a80>] (__wait_on_bit+0x54/0xa0)
[<c0769a80>] (__wait_on_bit+0x54/0xa0) from [<c00ca88c>] (wait_on_page_bit+0x8c/0x9c)
[<c00ca88c>] (wait_on_page_bit+0x8c/0x9c) from [<c00ca954>] (filemap_fdatawait_range+0x6c/0x110)
[<c00ca954>] (filemap_fdatawait_range+0x6c/0x110) from [<c00cbd8c>] (filemap_write_and_wait_range+0x54/0x78)
[<c00cbd8c>] (filemap_write_and_wait_range+0x54/0x78) from [<c0182e14>] (ext4_sync_file+0xdc/0x308)
[<c0182e14>] (ext4_sync_file+0xdc/0x308) from [<c01305f8>] (vfs_fsync_range+0x34/0x44)
[<c01305f8>] (vfs_fsync_range+0x34/0x44) from [<c01306b0>] (generic_write_sync+0x80/0x88)
[<c01306b0>] (generic_write_sync+0x80/0x88) from [<c00cc3fc>] (generic_file_aio_write+0xa0/0xb0)
[<c00cc3fc>] (generic_file_aio_write+0xa0/0xb0) from [<c0109198>] (do_sync_write+0x74/0x98)
[<c0109198>] (do_sync_write+0x74/0x98) from [<c0109ab8>] (vfs_write+0xd4/0x16c)
[<c0109ab8>] (vfs_write+0xd4/0x16c) from [<c0109df8>] (SyS_write+0x3c/0x60)
[<c0109df8>] (SyS_write+0x3c/0x60) from [<c000da00>] (ret_fast_syscall+0x0/0x30)

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Tested-by: Meiyou Chen <cmy@rock-chips.com>

diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c
index 33bdf60c8096a38aaf48743a70e8d9d7ea243f4d..05b8eaf06d7267349d89479bc7f1a1ddc19c5aaa 100755 (executable)
--- a/drivers/mmc/core/core.c
+++ b/drivers/mmc/core/core.c
@@ -396,6 +396,33 @@ static int __mmc_start_req(struct mmc_host *host, struct mmc_request *mrq)
        return 0;
 }
 
+static void mmc_get_req_timeout(struct mmc_request *mrq, u32 *timeout)
+{
+       if (!mrq->cmd->data) {
+               if (mrq->cmd->opcode == MMC_ERASE ||
+                   (mrq->cmd->opcode == MMC_ERASE_GROUP_START) ||
+                   (mrq->cmd->opcode == MMC_ERASE_GROUP_END) ||
+                   (mrq->cmd->opcode == MMC_SEND_STATUS))
+                       *timeout = 2500000;
+               else
+                       *timeout = 500;
+       } else {
+               *timeout = mrq->cmd->data->blocks *
+                       mrq->cmd->data->blksz * 500;
+               *timeout = (*timeout) ? (*timeout) : 1000;
+               if (*timeout > 8000)
+                       *timeout = 8000;
+       }
+
+       if ((mrq->cmd->opcode == SD_IO_RW_DIRECT) ||
+           (mrq->cmd->opcode == SD_IO_RW_EXTENDED))
+               *timeout = 8000;
+       else if ((mrq->cmd->opcode == MMC_SEND_TUNING_BLOCK_HS200) ||
+                (mrq->cmd->opcode == MMC_SEND_TUNING_BLOCK))
+               *timeout = 30;
+}
+
+
 /*
  * mmc_wait_for_data_req_done() - wait for request completed
  * @host: MMC host to prepare the command.
@@ -415,11 +442,24 @@ static int mmc_wait_for_data_req_done(struct mmc_host *host,
        struct mmc_context_info *context_info = &host->context_info;
        int err;
        unsigned long flags;
+       u32 timeout = 0;
+
+       mmc_get_req_timeout(mrq, &timeout);
 
        while (1) {
-               wait_event_interruptible(context_info->wait,
+               if (!wait_event_interruptible_timeout(context_info->wait,
                                         (context_info->is_done_rcv ||
-                                        context_info->is_new_req));
+                                        context_info->is_new_req),
+                                        msecs_to_jiffies(timeout))) {
+                       cmd = mrq->cmd;
+                       cmd->error = -ETIMEDOUT;
+                       host->ops->post_tmo(host);
+                       context_info->is_done_rcv = true;
+                       dev_err(mmc_dev(host),
+                               "req failed (CMD%u): error = %d, timeout = %dms\n",
+                               cmd->opcode, cmd->error, timeout);
+               }
+
                spin_lock_irqsave(&context_info->lock, flags);
                context_info->is_waiting_last_req = false;
                spin_unlock_irqrestore(&context_info->lock, flags);