UBIFS assert failed in ubifs_log_start_commit at 412 (pid 6)

UBIFS assert failed in ubifs_log_start_commit at 412 (pid 6)

[Richard Weinberger]

Hi!

Markus is facing this assert:

UBIFS assert failed in ubifs_log_start_commit at 412 (pid 6)
CPU: 0 PID: 6 Comm: kworker/u2:0 Not tainted 4.1.0-20150730-1 #1
Hardware name: Freescale i.MX53 (Device Tree Support)
Workqueue: writeback bdi_writeback_workfn (flush-ubifs_0_1)
[<c000fe89>] (unwind_backtrace) from [<c000ed8b>] (show_stack+0xb/0xc)
[<c000ed8b>] (show_stack) from [<c00e9327>] (ubifs_log_start_commit+0xff/0x264)
[<c00e9327>] (ubifs_log_start_commit) from [<c00e9cc5>] (do_commit+0x145/0x3a0)
[<c00e9cc5>] (do_commit) from [<c00db4ad>] (make_reservation+0x211/0x29c)
[<c00db4ad>] (make_reservation) from [<c00dbb4f>] (ubifs_jnl_write_data+0xfb/0x1c0)
[<c00dbb4f>] (ubifs_jnl_write_data) from [<c00dcd2f>] (do_writepage+0x8f/0x180)
[<c00dcd2f>] (do_writepage) from [<c0044ff3>] (__writepage+0xb/0x34)
[<c0044ff3>] (__writepage) from [<c0045fa9>] (write_cache_pages+0x165/0x1ea)
[<c0045fa9>] (write_cache_pages) from [<c004604f>] (generic_writepages+0x21/0x36)
[<c004604f>] (generic_writepages) from [<c007423d>] (__writeback_single_inode+0x25/0xd8)
[<c007423d>] (__writeback_single_inode) from [<c00744cd>] (writeback_sb_inodes+0x121/0x22c)
[<c00744cd>] (writeback_sb_inodes) from [<c0074619>] (__writeback_inodes_wb+0x41/0x68)
[<c0074619>] (__writeback_inodes_wb) from [<c00746df>] (wb_writeback+0x9f/0xe0)
[<c00746df>] (wb_writeback) from [<c0074853>] (bdi_writeback_workfn+0x133/0x208)
[<c0074853>] (bdi_writeback_workfn) from [<c00207d7>] (process_one_work+0x11b/0x1dc)
[<c00207d7>] (process_one_work) from [<c0020a63>] (worker_thread+0x1b1/0x28e)
[<c0020a63>] (worker_thread) from [<c0023263>] (kthread+0x87/0x98)
[<c0023263>] (kthread) from [<c000cde1>] (ret_from_fork+0x11/0x30)

Hu, this ubifs_assert was added by your commit 25601a3c9737fed554169759582c690b98ead5d4.
If I understand it correctly the assert is harmless as UBIFS can deal with that condition just fine.
The assertion was added to identify code paths where this can be triggered as you thought
you have fixed all paths?

IMHO the ubifs_assert() should be converted to a plain ubifs_msg().

Thanks,
//richard