Tool/software: Linux
We're running GStreamer on a TDA2E dual-core system using 4.1 Linux kernel. There are persistent problems with graphics buffer allocations resulting in this message:
[ 6725.856093] ===============================
[ 6725.860819] [ INFO: suspicious RCU usage. ]
[ 6725.865089] 4.1.13-rt11+ #5 Tainted: G O
[ 6725.870538] -------------------------------
[ 6725.875006] include/linux/fdtable.h:88 suspicious rcu_dereference_check() usage!
[ 6725.882662]
other info that might help us debug this:
[ 6725.890957]
rcu_scheduler_active = 1, debug_locks = 0
[ 6725.897548] no locks held by gst-launch-1.0/7211.
[ 6725.902634]
stack backtrace:
[ 6725.907258] CPU: 1 PID: 7211 Comm: gst-launch-1.0 Tainted: G O 4.1.13-rt11+ #5
[ 6725.915695] Hardware name: Generic DRA74X (Flattened Device Tree)
[ 6725.921871] [<c00197b8>] (unwind_backtrace) from [<c0015234>] (show_stack+0x20/0x24)
[ 6725.929700] [<c0015234>] (show_stack) from [<c07b0470>] (dump_stack+0x8c/0xcc)
[ 6725.937003] [<c07b0470>] (dump_stack) from [<c0099ea4>] (lockdep_rcu_suspicious+0xe8/0x11c)
[ 6725.945440] [<c0099ea4>] (lockdep_rcu_suspicious) from [<c067168c>] (rppc_ioctl+0x3a4/0x66c)
[ 6725.953963] [<c067168c>] (rppc_ioctl) from [<c019d26c>] (do_vfs_ioctl+0x450/0x6e0)
[ 6725.961608] [<c019d26c>] (do_vfs_ioctl) from [<c019d578>] (SyS_ioctl+0x7c/0x8c)
[ 6725.968996] [<c019d578>] (SyS_ioctl) from [<c0010980>] (ret_fast_syscall+0x0/0x54)
When this message appears, Gstreamer dies and reports
GST_ERROR_SYSTEM gstelement.c:1835:gst_element_message_full:<udpsrc0> posting message: Internal data flow error.
When I turn on Linux's dynamic debug facility for the file drivers/rpmsg/rpmsg_rpc.c, I see normal-looking communications via rpcc_ioctl between what I guess is IVAHD and the SOC. I cannot catch this error when the verbose debugging is turned on.
I haven't found any reports about a similar bug upstream. Thanks for any suggestions about how to address this problem.
Alison Chaiken
Peloton Technology
Mountain View, CA
