New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

vc4: Adafruit 3.5" SPI panel not supported #10

Open
anholt opened this Issue Feb 15, 2016 · 17 comments

Comments

Projects
None yet
4 participants
@anholt
Owner

anholt commented Feb 15, 2016

These are going to be tricker than the other panels: We need to use the transposer to write back into memory, then when the transposer is done we need to use the DMA engine to stream the new frame over SPI to the device. I hope.

anholt pushed a commit that referenced this issue Feb 27, 2016

Adrian Hunter Arnaldo Carvalho de Melo
perf tools: tracepoint_error() can receive e=NULL, robustify it
Fixes segmentation fault using, for instance:

  (gdb) run record -I -e intel_pt/tsc=1,noretcomp=1/u /bin/ls
  Starting program: /home/acme/bin/perf record -I -e intel_pt/tsc=1,noretcomp=1/u /bin/ls
  Missing separate debuginfos, use: dnf debuginfo-install glibc-2.22-7.fc23.x86_64
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib64/libthread_db.so.1".

 Program received signal SIGSEGV, Segmentation fault.
  0 x00000000004b9ea5 in tracepoint_error (e=0x0, err=13, sys=0x19b1370 "sched", name=0x19a5d00 "sched_switch") at util/parse-events.c:410
  (gdb) bt
  #0  0x00000000004b9ea5 in tracepoint_error (e=0x0, err=13, sys=0x19b1370 "sched", name=0x19a5d00 "sched_switch") at util/parse-events.c:410
  #1  0x00000000004b9fc5 in add_tracepoint (list=0x19a5d20, idx=0x7fffffffb8c0, sys_name=0x19b1370 "sched", evt_name=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:433
  #2  0x00000000004ba334 in add_tracepoint_event (list=0x19a5d20, idx=0x7fffffffb8c0, sys_name=0x19b1370 "sched", evt_name=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:498
  #3  0x00000000004bb699 in parse_events_add_tracepoint (list=0x19a5d20, idx=0x7fffffffb8c0, sys=0x19b1370 "sched", event=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:936
  #4  0x00000000004f6eda in parse_events_parse (_data=0x7fffffffb8b0, scanner=0x19a49d0) at util/parse-events.y:391
  #5  0x00000000004bc8e5 in parse_events__scanner (str=0x663ff2 "sched:sched_switch", data=0x7fffffffb8b0, start_token=258) at util/parse-events.c:1361
  #6  0x00000000004bca57 in parse_events (evlist=0x19a5220, str=0x663ff2 "sched:sched_switch", err=0x0) at util/parse-events.c:1401
  #7  0x0000000000518d5f in perf_evlist__can_select_event (evlist=0x19a3b90, str=0x663ff2 "sched:sched_switch") at util/record.c:253
  #8  0x0000000000553c42 in intel_pt_track_switches (evlist=0x19a3b90) at arch/x86/util/intel-pt.c:364
  #9  0x00000000005549d1 in intel_pt_recording_options (itr=0x19a2c40, evlist=0x19a3b90, opts=0x8edf68 <record+232>) at arch/x86/util/intel-pt.c:664
  #10 0x000000000051e076 in auxtrace_record__options (itr=0x19a2c40, evlist=0x19a3b90, opts=0x8edf68 <record+232>) at util/auxtrace.c:539
  #11 0x0000000000433368 in cmd_record (argc=1, argv=0x7fffffffde60, prefix=0x0) at builtin-record.c:1264
  #12 0x000000000049bec2 in run_builtin (p=0x8fa2a8 <commands+168>, argc=5, argv=0x7fffffffde60) at perf.c:390
  #13 0x000000000049c12a in handle_internal_command (argc=5, argv=0x7fffffffde60) at perf.c:451
  #14 0x000000000049c278 in run_argv (argcp=0x7fffffffdcbc, argv=0x7fffffffdcb0) at perf.c:495
  #15 0x000000000049c60a in main (argc=5, argv=0x7fffffffde60) at perf.c:618
(gdb)

Intel PT attempts to find the sched:sched_switch tracepoint but that seg
faults if tracefs is not readable, because the error reporting structure
is null, as errors are not reported when automatically adding
tracepoints.  Fix by checking before using.

Committer note:

This doesn't take place in a kernel that supports
perf_event_attr.context_switch, that is the default way that will be
used for tracking context switches, only in older kernels, like 4.2, in
a machine with Intel PT (e.g. Broadwell) for non-priviledged users.

Further info from a similar patch by Wang:

The error is in tracepoint_error: it assumes the 'e' parameter is valid.

However, there are many situation a parse_event() can be called without
parse_events_error. See result of

  $ grep 'parse_events(.*NULL)' ./tools/perf/ -r'

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Tong Zhang <ztong@vt.edu>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: stable@vger.kernel.org # v4.4+
Fixes: 1965817 ("perf tools: Enhance parsing events tracepoint error output")
Link: http://lkml.kernel.org/r/1453809921-24596-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
@anholt

This comment has been minimized.

Owner

anholt commented Mar 25, 2016

This will probably be supported in DRM prime mode with Noralf's tinydrm work.

anholt pushed a commit that referenced this issue Apr 20, 2016

perf tools: tracepoint_error() can receive e=NULL, robustify it
commit ec183d2 upstream.

Fixes segmentation fault using, for instance:

  (gdb) run record -I -e intel_pt/tsc=1,noretcomp=1/u /bin/ls
  Starting program: /home/acme/bin/perf record -I -e intel_pt/tsc=1,noretcomp=1/u /bin/ls
  Missing separate debuginfos, use: dnf debuginfo-install glibc-2.22-7.fc23.x86_64
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib64/libthread_db.so.1".

 Program received signal SIGSEGV, Segmentation fault.
  0 x00000000004b9ea5 in tracepoint_error (e=0x0, err=13, sys=0x19b1370 "sched", name=0x19a5d00 "sched_switch") at util/parse-events.c:410
  (gdb) bt
  #0  0x00000000004b9ea5 in tracepoint_error (e=0x0, err=13, sys=0x19b1370 "sched", name=0x19a5d00 "sched_switch") at util/parse-events.c:410
  #1  0x00000000004b9fc5 in add_tracepoint (list=0x19a5d20, idx=0x7fffffffb8c0, sys_name=0x19b1370 "sched", evt_name=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:433
  #2  0x00000000004ba334 in add_tracepoint_event (list=0x19a5d20, idx=0x7fffffffb8c0, sys_name=0x19b1370 "sched", evt_name=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:498
  #3  0x00000000004bb699 in parse_events_add_tracepoint (list=0x19a5d20, idx=0x7fffffffb8c0, sys=0x19b1370 "sched", event=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:936
  #4  0x00000000004f6eda in parse_events_parse (_data=0x7fffffffb8b0, scanner=0x19a49d0) at util/parse-events.y:391
  #5  0x00000000004bc8e5 in parse_events__scanner (str=0x663ff2 "sched:sched_switch", data=0x7fffffffb8b0, start_token=258) at util/parse-events.c:1361
  #6  0x00000000004bca57 in parse_events (evlist=0x19a5220, str=0x663ff2 "sched:sched_switch", err=0x0) at util/parse-events.c:1401
  #7  0x0000000000518d5f in perf_evlist__can_select_event (evlist=0x19a3b90, str=0x663ff2 "sched:sched_switch") at util/record.c:253
  #8  0x0000000000553c42 in intel_pt_track_switches (evlist=0x19a3b90) at arch/x86/util/intel-pt.c:364
  #9  0x00000000005549d1 in intel_pt_recording_options (itr=0x19a2c40, evlist=0x19a3b90, opts=0x8edf68 <record+232>) at arch/x86/util/intel-pt.c:664
  #10 0x000000000051e076 in auxtrace_record__options (itr=0x19a2c40, evlist=0x19a3b90, opts=0x8edf68 <record+232>) at util/auxtrace.c:539
  #11 0x0000000000433368 in cmd_record (argc=1, argv=0x7fffffffde60, prefix=0x0) at builtin-record.c:1264
  #12 0x000000000049bec2 in run_builtin (p=0x8fa2a8 <commands+168>, argc=5, argv=0x7fffffffde60) at perf.c:390
  #13 0x000000000049c12a in handle_internal_command (argc=5, argv=0x7fffffffde60) at perf.c:451
  #14 0x000000000049c278 in run_argv (argcp=0x7fffffffdcbc, argv=0x7fffffffdcb0) at perf.c:495
  #15 0x000000000049c60a in main (argc=5, argv=0x7fffffffde60) at perf.c:618
(gdb)

Intel PT attempts to find the sched:sched_switch tracepoint but that seg
faults if tracefs is not readable, because the error reporting structure
is null, as errors are not reported when automatically adding
tracepoints.  Fix by checking before using.

Committer note:

This doesn't take place in a kernel that supports
perf_event_attr.context_switch, that is the default way that will be
used for tracking context switches, only in older kernels, like 4.2, in
a machine with Intel PT (e.g. Broadwell) for non-priviledged users.

Further info from a similar patch by Wang:

The error is in tracepoint_error: it assumes the 'e' parameter is valid.

However, there are many situation a parse_event() can be called without
parse_events_error. See result of

  $ grep 'parse_events(.*NULL)' ./tools/perf/ -r'

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Tong Zhang <ztong@vt.edu>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: 1965817 ("perf tools: Enhance parsing events tracepoint error output")
Link: http://lkml.kernel.org/r/1453809921-24596-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

anholt pushed a commit that referenced this issue May 6, 2016

hwmon: (max1111) Return -ENODEV from max1111_read_channel if not inst…
…antiated

commit 3c2e226 upstream.

arm:pxa_defconfig can result in the following crash if the max1111 driver
is not instantiated.

Unhandled fault: page domain fault (0x01b) at 0x00000000
pgd = c0004000
[00000000] *pgd=00000000
Internal error: : 1b [#1] PREEMPT ARM
Modules linked in:
CPU: 0 PID: 300 Comm: kworker/0:1 Not tainted 4.5.0-01301-g1701f680407c #10
Hardware name: SHARP Akita
Workqueue: events sharpsl_charge_toggle
task: c390a000 ti: c391e000 task.ti: c391e000
PC is at max1111_read_channel+0x20/0x30
LR is at sharpsl_pm_pxa_read_max1111+0x2c/0x3c
pc : [<c03aaab0>]    lr : [<c0024b50>]    psr: 20000013
...
[<c03aaab0>] (max1111_read_channel) from [<c0024b50>]
					(sharpsl_pm_pxa_read_max1111+0x2c/0x3c)
[<c0024b50>] (sharpsl_pm_pxa_read_max1111) from [<c00262e0>]
					(spitzpm_read_devdata+0x5c/0xc4)
[<c00262e0>] (spitzpm_read_devdata) from [<c0024094>]
					(sharpsl_check_battery_temp+0x78/0x110)
[<c0024094>] (sharpsl_check_battery_temp) from [<c0024f9c>]
					(sharpsl_charge_toggle+0x48/0x110)
[<c0024f9c>] (sharpsl_charge_toggle) from [<c004429c>]
					(process_one_work+0x14c/0x48c)
[<c004429c>] (process_one_work) from [<c0044618>] (worker_thread+0x3c/0x5d4)
[<c0044618>] (worker_thread) from [<c004a238>] (kthread+0xd0/0xec)
[<c004a238>] (kthread) from [<c000a670>] (ret_from_fork+0x14/0x24)

This can occur because the SPI controller driver (SPI_PXA2XX) is built as
module and thus not necessarily loaded. While building SPI_PXA2XX into the
kernel would make the problem disappear, it appears prudent to ensure that
the driver is instantiated before accessing its data structures.

Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

anholt pushed a commit that referenced this issue Oct 17, 2016

IB/core: correctly handle rdma_rw_init_mrs() failure
Function ib_create_qp() was failing to return an error when
rdma_rw_init_mrs() fails, causing a crash further down in ib_create_qp()
when trying to dereferece the qp pointer which was actually a negative
errno.

The crash:

crash> log|grep BUG
[  136.458121] BUG: unable to handle kernel NULL pointer dereference at 0000000000000098
crash> bt
PID: 3736   TASK: ffff8808543215c0  CPU: 2   COMMAND: "kworker/u64:2"
 #0 [ffff88084d323340] machine_kexec at ffffffff8105fbb0
 #1 [ffff88084d3233b0] __crash_kexec at ffffffff81116758
 #2 [ffff88084d323480] crash_kexec at ffffffff8111682d
 #3 [ffff88084d3234b0] oops_end at ffffffff81032bd6
 #4 [ffff88084d3234e0] no_context at ffffffff8106e431
 #5 [ffff88084d323530] __bad_area_nosemaphore at ffffffff8106e610
 #6 [ffff88084d323590] bad_area_nosemaphore at ffffffff8106e6f4
 #7 [ffff88084d3235a0] __do_page_fault at ffffffff8106ebdc
 #8 [ffff88084d323620] do_page_fault at ffffffff8106f057
 #9 [ffff88084d323660] page_fault at ffffffff816e3148
    [exception RIP: ib_create_qp+427]
    RIP: ffffffffa02554fb  RSP: ffff88084d323718  RFLAGS: 00010246
    RAX: 0000000000000004  RBX: fffffffffffffff4  RCX: 000000018020001f
    RDX: ffff880830997fc0  RSI: 0000000000000001  RDI: ffff88085f407200
    RBP: ffff88084d323778   R8: 0000000000000001   R9: ffffea0020bae210
    R10: ffffea0020bae218  R11: 0000000000000001  R12: ffff88084d3237c8
    R13: 00000000fffffff4  R14: ffff880859fa5000  R15: ffff88082eb89800
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
#10 [ffff88084d323780] rdma_create_qp at ffffffffa0782681 [rdma_cm]
#11 [ffff88084d3237b0] nvmet_rdma_create_queue_ib at ffffffffa07c43f3 [nvmet_rdma]
#12 [ffff88084d323860] nvmet_rdma_alloc_queue at ffffffffa07c5ba9 [nvmet_rdma]
#13 [ffff88084d323900] nvmet_rdma_queue_connect at ffffffffa07c5c96 [nvmet_rdma]
#14 [ffff88084d323980] nvmet_rdma_cm_handler at ffffffffa07c6450 [nvmet_rdma]
#15 [ffff88084d3239b0] iw_conn_req_handler at ffffffffa0787480 [rdma_cm]
#16 [ffff88084d323a60] cm_conn_req_handler at ffffffffa0775f06 [iw_cm]
#17 [ffff88084d323ab0] process_event at ffffffffa0776019 [iw_cm]
#18 [ffff88084d323af0] cm_work_handler at ffffffffa0776170 [iw_cm]
#19 [ffff88084d323cb0] process_one_work at ffffffff810a1483
#20 [ffff88084d323d90] worker_thread at ffffffff810a211d
#21 [ffff88084d323ec0] kthread at ffffffff810a6c5c
#22 [ffff88084d323f50] ret_from_fork at ffffffff816e1ebf

Fixes: 632bc3f ("IB/core, RDMA RW API: Do not exceed QP SGE send limit")
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Cc: stable@vger.kernel.org
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
@notro

This comment has been minimized.

notro commented Jan 16, 2017

I tried to offload rendering to vc4 when using tinydrm, but I couldn't get it to work.
Should it work?

$ sudo lightdm ^Z

# working fine
$ DISPLAY=:0 glxgears -fullscreen

$ DISPLAY=:0 glxinfo | grep "OpenGL renderer"
OpenGL renderer string: Software Rasterizer

$ DISPLAY=:0 xrandr --listproviders
Providers: number : 2
Provider 0: id: 0x65 cap: 0x2, Sink Output crtcs: 1 outputs: 1 associated providers: 0 name:modesetting
Provider 1: id: 0x43 cap: 0x2, Sink Output crtcs: 3 outputs: 1 associated providers: 0 name:modesetting

$ DISPLAY=:0 xrandr --setprovideroffloadsink 1 0
X Error of failed request:  BadValue (integer parameter out of range for operation)
  Major opcode of failed request:  140 (RANDR)
  Minor opcode of failed request:  34 ()
  Value in failed request:  0x43
  Serial number of failed request:  16
  Current serial number in output stream:  17

$ dmesg
[ 3286.999925] [drm:drm_helper_probe_single_connector_modes] [CONNECTOR:25:Virtual-1]
[ 3286.999965] [drm:drm_helper_probe_single_connector_modes] [CONNECTOR:25:Virtual-1] probed modes :
[ 3286.999985] [drm:drm_mode_debug_printmodeline] Modeline 29:"320x240" 0 1 320 320 320 320 240 240 240 240 0x48 0x0
[ 3287.000884] [drm:drm_helper_probe_single_connector_modes] [CONNECTOR:25:Virtual-1]
[ 3287.000904] [drm:drm_helper_probe_single_connector_modes] [CONNECTOR:25:Virtual-1] probed modes :
[ 3287.000921] [drm:drm_mode_debug_printmodeline] Modeline 29:"320x240" 0 1 320 320 320 320 240 240 240 240 0x48 0x0

/var/log/Xorg.0.log

@anholt

This comment has been minimized.

Owner

anholt commented Jan 17, 2017

Thanks for looking into this! Here's my output, without tinydrm:

# xrandr --listproviders
Providers: number : 1
Provider 0: id: 0x43 cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 3 outputs: 1 associated providers: 0 name:modesetting

So you don't have export or offload listed. glamor acceleration is required for PRIME, which your Xorg.log said didn't get initialized. Do you maybe not have the userspace side of the driver installed?

@notro

This comment has been minimized.

notro commented Jan 17, 2017

I have now tried the latest Raspian 2017-01-11, same result.
Then I followed your guide: https://github.com/anholt/mesa/wiki/VC4-complete-Raspbian-upgrade

$ DISPLAY=:0 xrandr --listproviders
XIO:  fatal IO error 11 (Resource temporarily unavailable) on X server ":0"
      after 12 requests (12 known processed) with 0 events remaining.

This is the same error that I got on 1.16.4. Upgrading xserver-xorg to 1.18.4 gave me the result in my previous post.

/var/log/Xorg.0.log

If you don't see any obvious solution to this, I'll just pick it up again in a few months when it has matured some more.

anholt pushed a commit that referenced this issue Jan 23, 2017

xfs: fix up xfs_swap_extent_forks inline extent handling
commit 4dfce57 upstream.

There have been several reports over the years of NULL pointer
dereferences in xfs_trans_log_inode during xfs_fsr processes,
when the process is doing an fput and tearing down extents
on the temporary inode, something like:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
PID: 29439  TASK: ffff880550584fa0  CPU: 6   COMMAND: "xfs_fsr"
    [exception RIP: xfs_trans_log_inode+0x10]
 #9 [ffff8800a57bbbe0] xfs_bunmapi at ffffffffa037398e [xfs]
#10 [ffff8800a57bbce8] xfs_itruncate_extents at ffffffffa0391b29 [xfs]
#11 [ffff8800a57bbd88] xfs_inactive_truncate at ffffffffa0391d0c [xfs]
#12 [ffff8800a57bbdb8] xfs_inactive at ffffffffa0392508 [xfs]
#13 [ffff8800a57bbdd8] xfs_fs_evict_inode at ffffffffa035907e [xfs]
#14 [ffff8800a57bbe00] evict at ffffffff811e1b67
#15 [ffff8800a57bbe28] iput at ffffffff811e23a5
#16 [ffff8800a57bbe58] dentry_kill at ffffffff811dcfc8
#17 [ffff8800a57bbe88] dput at ffffffff811dd06c
#18 [ffff8800a57bbea8] __fput at ffffffff811c823b
#19 [ffff8800a57bbef0] ____fput at ffffffff811c846e
#20 [ffff8800a57bbf00] task_work_run at ffffffff81093b27
#21 [ffff8800a57bbf30] do_notify_resume at ffffffff81013b0c
#22 [ffff8800a57bbf50] int_signal at ffffffff8161405d

As it turns out, this is because the i_itemp pointer, along
with the d_ops pointer, has been overwritten with zeros
when we tear down the extents during truncate.  When the in-core
inode fork on the temporary inode used by xfs_fsr was originally
set up during the extent swap, we mistakenly looked at di_nextents
to determine whether all extents fit inline, but this misses extents
generated by speculative preallocation; we should be using if_bytes
instead.

This mistake corrupts the in-memory inode, and code in
xfs_iext_remove_inline eventually gets bad inputs, causing
it to memmove and memset incorrect ranges; this became apparent
because the two values in ifp->if_u2.if_inline_ext[1] contained
what should have been in d_ops and i_itemp; they were memmoved due
to incorrect array indexing and then the original locations
were zeroed with memset, again due to an array overrun.

Fix this by properly using i_df.if_bytes to determine the number
of extents, not di_nextents.

Thanks to dchinner for looking at this with me and spotting the
root cause.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

anholt pushed a commit that referenced this issue Mar 30, 2017

dccp/tcp: fix routing redirect race
[ Upstream commit 45caeaa ]

As Eric Dumazet pointed out this also needs to be fixed in IPv6.
v2: Contains the IPv6 tcp/Ipv6 dccp patches as well.

We have seen a few incidents lately where a dst_enty has been freed
with a dangling TCP socket reference (sk->sk_dst_cache) pointing to that
dst_entry. If the conditions/timings are right a crash then ensues when the
freed dst_entry is referenced later on. A Common crashing back trace is:

 #8 [] page_fault at ffffffff8163e648
    [exception RIP: __tcp_ack_snd_check+74]
.
.
 #9 [] tcp_rcv_established at ffffffff81580b64
#10 [] tcp_v4_do_rcv at ffffffff8158b54a
#11 [] tcp_v4_rcv at ffffffff8158cd02
#12 [] ip_local_deliver_finish at ffffffff815668f4
#13 [] ip_local_deliver at ffffffff81566bd9
#14 [] ip_rcv_finish at ffffffff8156656d
#15 [] ip_rcv at ffffffff81566f06
#16 [] __netif_receive_skb_core at ffffffff8152b3a2
#17 [] __netif_receive_skb at ffffffff8152b608
#18 [] netif_receive_skb at ffffffff8152b690
#19 [] vmxnet3_rq_rx_complete at ffffffffa015eeaf [vmxnet3]
#20 [] vmxnet3_poll_rx_only at ffffffffa015f32a [vmxnet3]
#21 [] net_rx_action at ffffffff8152bac2
#22 [] __do_softirq at ffffffff81084b4f
#23 [] call_softirq at ffffffff8164845c
#24 [] do_softirq at ffffffff81016fc5
#25 [] irq_exit at ffffffff81084ee5
#26 [] do_IRQ at ffffffff81648ff8

Of course it may happen with other NIC drivers as well.

It's found the freed dst_entry here:

 224 static bool tcp_in_quickack_mode(struct sock *sk)↩
 225 {↩
 226 ▹       const struct inet_connection_sock *icsk = inet_csk(sk);↩
 227 ▹       const struct dst_entry *dst = __sk_dst_get(sk);↩
 228 ↩
 229 ▹       return (dst && dst_metric(dst, RTAX_QUICKACK)) ||↩
 230 ▹       ▹       (icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong);↩
 231 }↩

But there are other backtraces attributed to the same freed dst_entry in
netfilter code as well.

All the vmcores showed 2 significant clues:

- Remote hosts behind the default gateway had always been redirected to a
different gateway. A rtable/dst_entry will be added for that host. Making
more dst_entrys with lower reference counts. Making this more probable.

- All vmcores showed a postitive LockDroppedIcmps value, e.g:

LockDroppedIcmps                  267

A closer look at the tcp_v4_err() handler revealed that do_redirect() will run
regardless of whether user space has the socket locked. This can result in a
race condition where the same dst_entry cached in sk->sk_dst_entry can be
decremented twice for the same socket via:

do_redirect()->__sk_dst_check()-> dst_release().

Which leads to the dst_entry being prematurely freed with another socket
pointing to it via sk->sk_dst_cache and a subsequent crash.

To fix this skip do_redirect() if usespace has the socket locked. Instead let
the redirect take place later when user space does not have the socket
locked.

The dccp/IPv6 code is very similar in this respect, so fixing it there too.

As Eric Garver pointed out the following commit now invalidates routes. Which
can set the dst->obsolete flag so that ipv4_dst_check() returns null and
triggers the dst_release().

Fixes: ceb3320 ("ipv4: Kill routes during PMTU/redirect updates.")
Cc: Eric Garver <egarver@redhat.com>
Cc: Hannes Sowa <hsowa@redhat.com>
Signed-off-by: Jon Maxwell <jmaxwell37@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

anholt pushed a commit that referenced this issue Mar 30, 2017

dccp/tcp: fix routing redirect race
[ Upstream commit 45caeaa ]

As Eric Dumazet pointed out this also needs to be fixed in IPv6.
v2: Contains the IPv6 tcp/Ipv6 dccp patches as well.

We have seen a few incidents lately where a dst_enty has been freed
with a dangling TCP socket reference (sk->sk_dst_cache) pointing to that
dst_entry. If the conditions/timings are right a crash then ensues when the
freed dst_entry is referenced later on. A Common crashing back trace is:

 #8 [] page_fault at ffffffff8163e648
    [exception RIP: __tcp_ack_snd_check+74]
.
.
 #9 [] tcp_rcv_established at ffffffff81580b64
#10 [] tcp_v4_do_rcv at ffffffff8158b54a
#11 [] tcp_v4_rcv at ffffffff8158cd02
#12 [] ip_local_deliver_finish at ffffffff815668f4
#13 [] ip_local_deliver at ffffffff81566bd9
#14 [] ip_rcv_finish at ffffffff8156656d
#15 [] ip_rcv at ffffffff81566f06
#16 [] __netif_receive_skb_core at ffffffff8152b3a2
#17 [] __netif_receive_skb at ffffffff8152b608
#18 [] netif_receive_skb at ffffffff8152b690
#19 [] vmxnet3_rq_rx_complete at ffffffffa015eeaf [vmxnet3]
#20 [] vmxnet3_poll_rx_only at ffffffffa015f32a [vmxnet3]
#21 [] net_rx_action at ffffffff8152bac2
#22 [] __do_softirq at ffffffff81084b4f
#23 [] call_softirq at ffffffff8164845c
#24 [] do_softirq at ffffffff81016fc5
#25 [] irq_exit at ffffffff81084ee5
#26 [] do_IRQ at ffffffff81648ff8

Of course it may happen with other NIC drivers as well.

It's found the freed dst_entry here:

 224 static bool tcp_in_quickack_mode(struct sock *sk)↩
 225 {↩
 226 ▹       const struct inet_connection_sock *icsk = inet_csk(sk);↩
 227 ▹       const struct dst_entry *dst = __sk_dst_get(sk);↩
 228 ↩
 229 ▹       return (dst && dst_metric(dst, RTAX_QUICKACK)) ||↩
 230 ▹       ▹       (icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong);↩
 231 }↩

But there are other backtraces attributed to the same freed dst_entry in
netfilter code as well.

All the vmcores showed 2 significant clues:

- Remote hosts behind the default gateway had always been redirected to a
different gateway. A rtable/dst_entry will be added for that host. Making
more dst_entrys with lower reference counts. Making this more probable.

- All vmcores showed a postitive LockDroppedIcmps value, e.g:

LockDroppedIcmps                  267

A closer look at the tcp_v4_err() handler revealed that do_redirect() will run
regardless of whether user space has the socket locked. This can result in a
race condition where the same dst_entry cached in sk->sk_dst_entry can be
decremented twice for the same socket via:

do_redirect()->__sk_dst_check()-> dst_release().

Which leads to the dst_entry being prematurely freed with another socket
pointing to it via sk->sk_dst_cache and a subsequent crash.

To fix this skip do_redirect() if usespace has the socket locked. Instead let
the redirect take place later when user space does not have the socket
locked.

The dccp/IPv6 code is very similar in this respect, so fixing it there too.

As Eric Garver pointed out the following commit now invalidates routes. Which
can set the dst->obsolete flag so that ipv4_dst_check() returns null and
triggers the dst_release().

Fixes: ceb3320 ("ipv4: Kill routes during PMTU/redirect updates.")
Cc: Eric Garver <egarver@redhat.com>
Cc: Hannes Sowa <hsowa@redhat.com>
Signed-off-by: Jon Maxwell <jmaxwell37@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

anholt pushed a commit that referenced this issue Apr 8, 2017

dccp/tcp: fix routing redirect race
As Eric Dumazet pointed out this also needs to be fixed in IPv6.
v2: Contains the IPv6 tcp/Ipv6 dccp patches as well.

We have seen a few incidents lately where a dst_enty has been freed
with a dangling TCP socket reference (sk->sk_dst_cache) pointing to that
dst_entry. If the conditions/timings are right a crash then ensues when the
freed dst_entry is referenced later on. A Common crashing back trace is:

 #8 [] page_fault at ffffffff8163e648
    [exception RIP: __tcp_ack_snd_check+74]
.
.
 #9 [] tcp_rcv_established at ffffffff81580b64
#10 [] tcp_v4_do_rcv at ffffffff8158b54a
#11 [] tcp_v4_rcv at ffffffff8158cd02
#12 [] ip_local_deliver_finish at ffffffff815668f4
#13 [] ip_local_deliver at ffffffff81566bd9
#14 [] ip_rcv_finish at ffffffff8156656d
#15 [] ip_rcv at ffffffff81566f06
#16 [] __netif_receive_skb_core at ffffffff8152b3a2
#17 [] __netif_receive_skb at ffffffff8152b608
#18 [] netif_receive_skb at ffffffff8152b690
#19 [] vmxnet3_rq_rx_complete at ffffffffa015eeaf [vmxnet3]
#20 [] vmxnet3_poll_rx_only at ffffffffa015f32a [vmxnet3]
#21 [] net_rx_action at ffffffff8152bac2
#22 [] __do_softirq at ffffffff81084b4f
#23 [] call_softirq at ffffffff8164845c
#24 [] do_softirq at ffffffff81016fc5
#25 [] irq_exit at ffffffff81084ee5
#26 [] do_IRQ at ffffffff81648ff8

Of course it may happen with other NIC drivers as well.

It's found the freed dst_entry here:

 224 static bool tcp_in_quickack_mode(struct sock *sk)↩
 225 {↩
 226 ▹       const struct inet_connection_sock *icsk = inet_csk(sk);↩
 227 ▹       const struct dst_entry *dst = __sk_dst_get(sk);↩
 228 ↩
 229 ▹       return (dst && dst_metric(dst, RTAX_QUICKACK)) ||↩
 230 ▹       ▹       (icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong);↩
 231 }↩

But there are other backtraces attributed to the same freed dst_entry in
netfilter code as well.

All the vmcores showed 2 significant clues:

- Remote hosts behind the default gateway had always been redirected to a
different gateway. A rtable/dst_entry will be added for that host. Making
more dst_entrys with lower reference counts. Making this more probable.

- All vmcores showed a postitive LockDroppedIcmps value, e.g:

LockDroppedIcmps                  267

A closer look at the tcp_v4_err() handler revealed that do_redirect() will run
regardless of whether user space has the socket locked. This can result in a
race condition where the same dst_entry cached in sk->sk_dst_entry can be
decremented twice for the same socket via:

do_redirect()->__sk_dst_check()-> dst_release().

Which leads to the dst_entry being prematurely freed with another socket
pointing to it via sk->sk_dst_cache and a subsequent crash.

To fix this skip do_redirect() if usespace has the socket locked. Instead let
the redirect take place later when user space does not have the socket
locked.

The dccp/IPv6 code is very similar in this respect, so fixing it there too.

As Eric Garver pointed out the following commit now invalidates routes. Which
can set the dst->obsolete flag so that ipv4_dst_check() returns null and
triggers the dst_release().

Fixes: ceb3320 ("ipv4: Kill routes during PMTU/redirect updates.")
Cc: Eric Garver <egarver@redhat.com>
Cc: Hannes Sowa <hsowa@redhat.com>
Signed-off-by: Jon Maxwell <jmaxwell37@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

anholt pushed a commit that referenced this issue Apr 19, 2017

cpuidle: Validate cpu_dev in cpuidle_add_sysfs()
If a given cpu is not in cpu_present and cpu hotplug
is disabled, arch can skip setting up the cpu_dev.

Arch cpuidle driver should pass correct cpu mask
for registration, but failing to do so by the driver
causes error to propagate and crash like this:

[   30.076045] Unable to handle kernel paging request for data at address 0x00000048
[   30.076100] Faulting instruction address: 0xc0000000007b2f30
cpu 0x4d: Vector: 300 (Data Access) at [c000003feb18b670]
    pc: c0000000007b2f30: kobject_get+0x20/0x70
    lr: c0000000007b3c94: kobject_add_internal+0x54/0x3f0
    sp: c000003feb18b8f0
   msr: 9000000000009033
   dar: 48
 dsisr: 40000000
  current = 0xc000003fd2ed8300
  paca    = 0xc00000000fbab500   softe: 0        irq_happened: 0x01
    pid   = 1, comm = swapper/0
Linux version 4.11.0-rc2-svaidy+ (sv@sagarika) (gcc version 6.2.0
20161005 (Ubuntu 6.2.0-5ubuntu12) ) #10 SMP Sun Mar 19 00:08:09 IST 2017
enter ? for help
[c000003feb18b960] c0000000007b3c94 kobject_add_internal+0x54/0x3f0
[c000003feb18b9f0] c0000000007b43a4 kobject_init_and_add+0x64/0xa0
[c000003feb18ba70] c000000000e284f4 cpuidle_add_sysfs+0xb4/0x130
[c000003feb18baf0] c000000000e26038 cpuidle_register_device+0x118/0x1c0
[c000003feb18bb30] c000000000e26c48 cpuidle_register+0x78/0x120
[c000003feb18bbc0] c00000000168fd9c powernv_processor_idle_init+0x110/0x1c4
[c000003feb18bc40] c00000000000cff8 do_one_initcall+0x68/0x1d0
[c000003feb18bd00] c0000000016242f4 kernel_init_freeable+0x280/0x360
[c000003feb18bdc0] c00000000000d864 kernel_init+0x24/0x160
[c000003feb18be30] c00000000000b4e8 ret_from_kernel_thread+0x5c/0x74

Validating cpu_dev fixes the crash and reports correct error message like:

[   30.163506] Failed to register cpuidle device for cpu136
[   30.173329] Registration of powernv driver failed.

Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
[ rjw: Comment massage ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

anholt pushed a commit that referenced this issue Jun 15, 2017

i2c: i2c-tiny-usb: fix buffer not being DMA capable
commit 5165da5 upstream.

Since v4.9 i2c-tiny-usb generates the below call trace
and longer works, since it can't communicate with the
USB device. The reason is, that since v4.9 the USB
stack checks, that the buffer it should transfer is DMA
capable. This was a requirement since v2.2 days, but it
usually worked nevertheless.

[   17.504959] ------------[ cut here ]------------
[   17.505488] WARNING: CPU: 0 PID: 93 at drivers/usb/core/hcd.c:1587 usb_hcd_map_urb_for_dma+0x37c/0x570
[   17.506545] transfer buffer not dma capable
[   17.507022] Modules linked in:
[   17.507370] CPU: 0 PID: 93 Comm: i2cdetect Not tainted 4.11.0-rc8+ #10
[   17.508103] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
[   17.509039] Call Trace:
[   17.509320]  ? dump_stack+0x5c/0x78
[   17.509714]  ? __warn+0xbe/0xe0
[   17.510073]  ? warn_slowpath_fmt+0x5a/0x80
[   17.510532]  ? nommu_map_sg+0xb0/0xb0
[   17.510949]  ? usb_hcd_map_urb_for_dma+0x37c/0x570
[   17.511482]  ? usb_hcd_submit_urb+0x336/0xab0
[   17.511976]  ? wait_for_completion_timeout+0x12f/0x1a0
[   17.512549]  ? wait_for_completion_timeout+0x65/0x1a0
[   17.513125]  ? usb_start_wait_urb+0x65/0x160
[   17.513604]  ? usb_control_msg+0xdc/0x130
[   17.514061]  ? usb_xfer+0xa4/0x2a0
[   17.514445]  ? __i2c_transfer+0x108/0x3c0
[   17.514899]  ? i2c_transfer+0x57/0xb0
[   17.515310]  ? i2c_smbus_xfer_emulated+0x12f/0x590
[   17.515851]  ? _raw_spin_unlock_irqrestore+0x11/0x20
[   17.516408]  ? i2c_smbus_xfer+0x125/0x330
[   17.516876]  ? i2c_smbus_xfer+0x125/0x330
[   17.517329]  ? i2cdev_ioctl_smbus+0x1c1/0x2b0
[   17.517824]  ? i2cdev_ioctl+0x75/0x1c0
[   17.518248]  ? do_vfs_ioctl+0x9f/0x600
[   17.518671]  ? vfs_write+0x144/0x190
[   17.519078]  ? SyS_ioctl+0x74/0x80
[   17.519463]  ? entry_SYSCALL_64_fastpath+0x1e/0xad
[   17.519959] ---[ end trace d047c04982f5ac50 ]---

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Till Harbaum <till@harbaum.org>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

@notro notro referenced this issue Jul 16, 2017

Merged

tinydrm backport #2119

anholt pushed a commit that referenced this issue Aug 15, 2017

mm: take memory hotplug lock within numa_zonelist_order_handler()
Andre Wild reported the following warning:

  WARNING: CPU: 2 PID: 1205 at kernel/cpu.c:240 lockdep_assert_cpus_held+0x4c/0x60
  Modules linked in:
  CPU: 2 PID: 1205 Comm: bash Not tainted 4.13.0-rc2-00022-gfd2b2c57ec20 #10
  Hardware name: IBM 2964 N96 702 (z/VM 6.4.0)
  task: 00000000701d8100 task.stack: 0000000073594000
  Krnl PSW : 0704f00180000000 0000000000145e24 (lockdep_assert_cpus_held+0x4c/0x60)
  ...
  Call Trace:
   lockdep_assert_cpus_held+0x42/0x60)
   stop_machine_cpuslocked+0x62/0xf0
   build_all_zonelists+0x92/0x150
   numa_zonelist_order_handler+0x102/0x150
   proc_sys_call_handler.isra.12+0xda/0x118
   proc_sys_write+0x34/0x48
   __vfs_write+0x3c/0x178
   vfs_write+0xbc/0x1a0
   SyS_write+0x66/0xc0
   system_call+0xc4/0x2b0
   locks held by bash/1205:
   #0:  (sb_writers#4){.+.+.+}, at: vfs_write+0xa6/0x1a0
   #1:  (zl_order_mutex){+.+...}, at: numa_zonelist_order_handler+0x44/0x150
   #2:  (zonelists_mutex){+.+...}, at: numa_zonelist_order_handler+0xf4/0x150
  Last Breaking-Event-Address:
    lockdep_assert_cpus_held+0x48/0x60

This can be easily triggered with e.g.

    echo n > /proc/sys/vm/numa_zonelist_order

In commit 3f906ba ("mm/memory-hotplug: switch locking to a percpu
rwsem") memory hotplug locking was changed to fix a potential deadlock.

This also switched the stop_machine() invocation within
build_all_zonelists() to stop_machine_cpuslocked() which now expects
that online cpus are locked when being called.

This assumption is not true if build_all_zonelists() is being called
from numa_zonelist_order_handler().

In order to fix this simply add a mem_hotplug_begin()/mem_hotplug_done()
pair to numa_zonelist_order_handler().

Link: http://lkml.kernel.org/r/20170726111738.38768-1-heiko.carstens@de.ibm.com
Fixes: 3f906ba ("mm/memory-hotplug: switch locking to a percpu rwsem")
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reported-by: Andre Wild <wild@linux.vnet.ibm.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
@notro

This comment has been minimized.

notro commented Sep 22, 2017

Hi Eric, I watched your XDC talk yesterday and was reminded of this issue.

Using 2017-09-07-raspbian-stretch I'm able to get further:
(I've backported tinydrm to rpi-4.9.y)

$ DISPLAY=:0 xrandr --setprovideroutputsource 1 0

$ DISPLAY=:0 xrandr --listproviders
Providers: number : 2
Provider 0: id: 0x45 cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 3 outputs: 3 associated providers: 1 name:modesetting
Provider 1: id: 0x28b cap: 0x2, Sink Output crtcs: 1 outputs: 1 associated providers: 1 name:modesetting

$ DISPLAY=:0 xrandr
Screen 0: minimum 320 x 200, current 1920 x 1080, maximum 2048 x 2048
HDMI-1 connected primary 1920x1080+0+0 (normal left inverted right x axis y axis) 521mm x 293mm
   1920x1080     60.00*+  50.00    59.94
   1680x1050     59.88
   1600x900      60.00
   1280x1024     75.02    60.02
   1440x900      59.90
   1280x800      59.91
   1152x864      75.00
   1280x720      60.00    50.00    59.94
   1024x768      75.03    70.07    60.00
   832x624       74.55
   800x600       72.19    75.00    60.32    56.25
   720x576       50.00
   720x480       60.00    59.94
   640x480       75.00    72.81    66.67    60.00    59.94
   720x400       70.08
Composite-1 unknown connection (normal left inverted right x axis y axis)
   720x480       62.69
DSI-1 disconnected (normal left inverted right x axis y axis)
Virtual-1-1 connected (normal left inverted right x axis y axis)
   320x240        0.01 +

$ sudo sh -c "echo 0x1f > /sys/module/drm/parameters/debug"
$ sudo dmesg -C
$ DISPLAY=:0 xrandr --output Virtual-1-1 --auto

I get the desktop on the display, but it doesn't refresh.

tinydrm flushes the framebuffer when the pipe is enabled, on PAGE_FLIP and on DIRTYFB.

Here's the kernel log from the above commands: https://gist.github.com/notro/fb8ec4418244ad2af8bd5fbba67434b6
It starts with a few DIRTYFB and then it stops.
I've added linebreaks around the flushing and it's triggers.

Any idea why the flushing stops?

Edit:
Display was off, I moved the mouse and it turned on with updated content. This wasn't suprising in itself since it flushes on enable, but once again there was a few DIRTYFB IOCTL's: https://gist.github.com/notro/8a1bdd09b6cc15217bb4a86f6282a2c9

@anholt

This comment has been minimized.

Owner

anholt commented Sep 22, 2017

The Xorg.log would probably be useful here, to see what it says about dirty tracking

@notro

This comment has been minimized.

notro commented Sep 23, 2017

I finally got it to work without knowing why. xrandr kept giving me errors and it resulted in the desktop switching from HDMI to TFT. When I tried today it suddenly worked.
Here's the log: https://gist.github.com/notro/cf8006ce379d134808672e2bb61ea9cf

I blacklist the driver mi0283qt so the desktop shows up on HDMI and not TFT. X configuration is a nightmare, even more so now that it's automatic, there's no config to adjust... So I load the driver after boot. Here's the whole ssh session: https://gist.github.com/notro/4d22e75a694a33496c3ff01e95e2e572

@gattis

This comment has been minimized.

gattis commented Oct 5, 2017

Thank you for all your work on this, anholt and notro, I really appreciate what you guys are doing. I'm wrestling with this issue as well. Starting an open source audio/visual synthesizer project with both HDMI out for visual synthesis and a control interface on a riverdi touch FT801-based SPI TFT. Running on a CM3 compute module, pretty much maxing out all the peripherals with additional 2x SPIs for ADC, DAC, plus i2s audio codec, plus bluetooth over uart and wifi over sdio. Rest of GPIO I used for modular synth gate inputs/outputs. Pretty much just have the SPI left for the touchscreen. DSI would have been nice but nobody seems to make a DSI touch panel in my size (3.5"). So I wrote a tinydrm driver based around notro's work (attached: rvtft801.txt dts.txt)

Started with downstream arm 32-bit and jessie, had it mostly working except for this issue of offloading GPU via prime buffers. But with swrast I was getting 40fps glxgears which is about all I can expect out of the SPI (until I add compression which the FT801 supports but will use more rpi cpu).

Updated to stretch 32-bit, all of a sudden frame rates drop in half (and this still get this bug). Turns out stretch is shipping the gallium llvmpipe software rasterizer which I'm guessing is in no way intended to perform well on a 32-bit arm (especially without asimd, etc). I compiled mesa manually using anholt's instructions and got back dri swrast with the right configure options, but still had this bug. I hadn't tried in the past few days, maybe it's fixed now with what notro is saying.

Anyway, tried going upstream to arm64 with stretch (and then also to arm64 arch linux to be able to get the latest mesa/xserver nicely packaged to me) to see if this works better with gallium vc4/llvmpipe, ported rest of my peripherals over, only to realize that my tinydrm driver no longer works because of the arm64 zone dma situation. dma_coerce_mask_and_coherent returns -ENOMEM, and if I continue anyway, the drm gem fbdev helper tells me it "failed to allocate buffer with size 155648", "Failed to set initial hw configuration.", and tinydrm register tells me "Failed to initialize fbdev: -12"

Happy to code and contribute whatever needs to be fixed myself, but where to go from here? Do I add the dma zone allocation to the spi driver (which is already using 2 dmas for tx/rx), or to a framebuffer driver (but looks like framebuffer is set to be deprecated upstream? https://elinux.org/RPi_Upstreaming). DMA can't be disabled, right? I would think communication between the GPU and ARM is necessarily DMA.

Seems like this all this migration upstream both to the newer drm platform and to arm64, while clearly better architecturally, leaves a bit of a major regression hole for SPI displays. Is there a plan for them, or were they just going to be accidentally marked deprecated? Would like to help make sure this doesn't happen.

@gattis

This comment has been minimized.

gattis commented Oct 5, 2017

Looks like a bunch of people were complaining on this thread about slower llvmpipe in stretch (https://www.raspberrypi.org/forums/viewtopic.php?t=191791), not sure if they realized it was because the gallium llvmpipe software rast is meant for x86 and llvm on arm 32-bit is not going to be good at all, but it turned childish and someone closed the thread.

@notro

This comment has been minimized.

notro commented Oct 6, 2017

spi devices are not dma devices, it's the spi controller device that does dma.
When I made tinydrm I used the cma helper because it was easy to use and I had little understanding of drm. This meant I had to make it look like the spi device could do dma by using dma_coerce_mask_and_coherent(). Now dma_alloc_wc() would succeed. The dma address returned was useless, but that didn't matter because the spi core used the streaming api anyway to get a new address (spi_map_buf()).
Why it fails on arm64, I don't know. Never tried it.

I'm currently in the process of moving away from the cma helper: https://lists.freedesktop.org/archives/dri-devel/2017-October/154048.html

@gattis

This comment has been minimized.

gattis commented Oct 11, 2017

amazing, thanks notro... when I apply your patches, and skip the dma_coerce_mask_and_coherent call, tinydrm is now working on arm64 patched against latest rpi-4.14.y / bcmrpi3_defconfig. tried it on a pitft+ 2.8" (using your mi0283qt driver) with and without your patches to confirm that's what fixed it. dma_coerce_mask_and_coherent still fails with -5 (as I think expected) but apparently we don't need it anymore.

sadly no framerate improvement with arm64 using gallium llvmpipe, it uses less CPU but I think it may be doing full screen updates and maxing out the SPI, unlike the good old dri swrast days.

but gpu offloading seems close! I don't get any more segfaults, "xrandr --setprovideroutputsource 1 0" returns successfully now. I just get "xrandr: Configure crtc 3 failed" from "xrandr --output Virtual-1-1 --auto". Xorg log shows "failed to add fb -22". might be down to xorg configuration at this point (god help us).

one question- will these prime buffers allow me to (in theory at least) render a GPU-accelerated 1920x1080 opengl canvas on HDMI, plus a GPU-accelerated 320x240 gui interface on tinydrm? Or is this strictly just going to work for mirroring?

@gattis

This comment has been minimized.

gattis commented Oct 30, 2017

Took another crack at this, got a little closer this time. FYI for anyone trying to get drm prime working with vc4, your life will get a lot easier if you write a small mesa gallium userspace driver instead of trying to convince the mesa loader to do what you want. Just copy/paste this one for the pl111 and swap in the name of your driver. Then have fun inserting it into the mesa source tree, build, and you should end up with an xxx_dri.so in your mesa dri drivers folder. With that, X seems to do the right thing without much pulling your hair out. We might have to use renderonly_create_kms_dumb_buffer_for_resource (whew) instead of renderonly_create_gpu_import_for_resource for tinydrm drivers. Was thinking I'd try to submit a generic tinydrm version upstream to mesa with a fallback to llvmpipe if vc4 isn't available, but since tinydrm drivers all have different names, I'm not sure how mesa would bind the right one because the matching seems solely based off the name of the spi driver. And it's a shame you can't build out-of-tree gallium drivers or at least I haven't found the headers required to.

I have a feeling this will work for 32-bit kernels. For me on 64-bit, this works up until the mesa code calls drmPrimeFDToHandle and ioctls the tinydrm (kernel) driver with DRM_IOCTL_PRIME_FD_TO_HANDLE. From there:

  • tinydrm kernel driver dispatches that to drm_gem_prime_fd_to_handle
  • drm_gem_prime_fd_to_handle calls back to tinydrm kernel driver's prime_import function, aka,
  • generic gem_prime_import, which successfully finds the vc4's dma_buf, but then then
  • unsuccessfully tries to call dma_buff_attach on it, which with some more callbacks (and callforths?) gets to
  • drm_gem_cma_prime_get_sg_table (from the vc4's dmabuf), which fails at
  • dma_get_sgtable, where it tries to get the dma_map_ops of the vc4's bus
  • that fails further down when it tries to call arm64's get_arch_dma_ops and ends up in what's supposed to be an impossible code path and returns something called dummy_dma_ops

The comment in the source code at that point is:
/*

  • We expect no ISA devices, and all other DMA masters are expected to
  • have someone call arch_setup_dma_ops at device creation time.
    */

The (32-bit) arm code, on the other hand passes back the proper iommu api ops. So maybe the bcm2835-dma driver is supposed to be doing something different on arm64 than it does on arm? Or am I just compiling the kernel with the wrong options? It's weird because other devices seem to dma fine and i get boot messages from the dma api saying everything is fine. I'm guessing this is the same reason I couldn't attach my tinydrm driver to the CMA gem helper but was able to with notro's new vmalloc bo patches on arm64

@lategoodbye

This comment has been minimized.

Collaborator

lategoodbye commented Oct 30, 2017

Did you tried mainline kernel and its defconfig?

@gattis

This comment has been minimized.

gattis commented Oct 30, 2017

there is no arm64 mainline defconfig cm3 dts. a modified rpi3 dts plus some love gets you booting (if i remember, i needed to make a new dt-blob.bin and jump through some u-boot hoops). even though it's the same processor as the rpi 3, handful of things are different in the peripheral wiring and different gpio expanders etc.

I winded up basing off raspberrypi/4.14.y, made a custom dts, and i do merges from there first and then from mainline to get better arm64 drivers, and now also merging in drm-tip and hand patching in notro's work from dri-devel mailing list so I can get this kernel driver working on arm64. Adding 3 other kernel modules on top of that, one of which is an alsa-soc codec, and I think I need the legacy 2709-dma driver for that because if it's circular abilities or whatever, but it seems to work in cooperation with the bcm2835 dma driver fine and I'm doing SPI dma just fine with this same display driver. I'm too far deep in this thing to look back, so forging ahead, but hoping to meet up with you guys somewhere in the mainline arm64 bcm2837 4.14 LTS promised land. That probably means I'm on my own too unless this is broken for anyone else.

It's possible i have something wrong around the IOMMU / CMA / DMA coherency bits of the kernel config or my dts needs some kind of iommu entry on arm64, but i dont see any references to this stuff in mainline or rpi3 dts files. I haven't fully wrapped my head around it and maybe there's a bad combo mixing mainline and downstream drivers. but i've scoured and i don't see any evidence of anyone getting an spi display working on arm64 mainline. i've tried probably 10 different reasonable-looking config permutations, i was ratelimited by the fact that DMA config changes trigger a full kernel rebuild. but i just got a jetson tx2 and nvme ssd, now I'm smoking through arm64 native kernel builds in like 10 mins. I'm coming back at this soon, for now I think using vintage mesa sw rast is better for me because it sends partial screen updates and seems like even if we get prime working it's gonna be a full scanout (correct if i'm wrong) and our bottleneck is spi clock. though that will probably change for me when I add compression.

@notro

This comment has been minimized.

notro commented Oct 31, 2017

seems like even if we get prime working it's gonna be a full scanout

This depends on usespace which I know hardly anything about.
tinydrm does a full scanout on page flips, so if userspace does page flips and dirtyfb ioctls, we're in a bad shape. My brief testing with X and prime indicated that this might be the case.
Anyway I have to make sure that tinydrm doesn't scanout page flips if dirtyfb has been used. I haven't looked into this because there was some talk about doing dirtyfb through atomic modesetting (which handles page flips), so I'm waiting to see if something comes out that.

@gattis

This comment has been minimized.

gattis commented Nov 10, 2017

I tried this out on rpi-4.14.y 32-bit without any patches and indeed it worked, quite well. You just have to make the tinydrm gallium driver as I mentioned above. Every app I tried works much better on the spi display with gpu rendering. To answer my own questions, yes you can render on HDMI and tinydrm display at the same time. And the rendering is done with partial updates both in X and wayland (though not tested in raspbian, this is using fedora rawhide), and there is way less spi traffic and frame rates can get really high. All the xrandr prime commands seem to work. Thanks again for everyone's work! Shortcut method for building your own gallium driver:

unpack or git clone mesa source, then run these commands, substituting ft8xx with the name of your tinydrm driver (the name you use with modprobe):

find . -type f -print0 | xargs -0 sed -i 's/pl111/ft8xx/g'
find . -type f -print0 | xargs -0 sed -i 's/PL111/FT8XX/g'
mv src/gallium/drivers/pl111 src/gallium/drivers/ft8xx
mv src/gallium/winsys/pl111 src/gallium/winsys/ft8xx
mv src/gallium/winsys/ft8xx/drm/pl111_drm_public.h src/gallium/winsys/ft8xx/drm/ft8xx_drm_public.h
mv src/gallium/winsys/ft8xx/drm/pl111_drm_winsys.c src/gallium/winsys/ft8xx/drm/ft8xx_drm_winsys.c

configure making sure you add --with-gallium-drivers=vc4,ft8xx

make

anholt pushed a commit that referenced this issue Nov 21, 2017

perf buildid-list: Fix crash when processing PERF_RECORD_NAMESPACE
Thomas reported that 'perf buildid-list' gets a SEGFAULT due to NULL
pointer deref when he ran it on a data with namespace events.  It was
because the buildid_id__mark_dso_hit_ops lacks the namespace event
handler and perf_too__fill_default() didn't set it.

  Program received signal SIGSEGV, Segmentation fault.
  0x0000000000000000 in ?? ()
  Missing separate debuginfos, use: dnf debuginfo-install audit-libs-2.7.7-1.fc25.s390x bzip2-libs-1.0.6-21.fc25.s390x elfutils-libelf-0.169-1.fc25.s390x
  +elfutils-libs-0.169-1.fc25.s390x libcap-ng-0.7.8-1.fc25.s390x numactl-libs-2.0.11-2.ibm.fc25.s390x openssl-libs-1.1.0e-1.1.ibm.fc25.s390x perl-libs-5.24.1-386.fc25.s390x
  +python-libs-2.7.13-2.fc25.s390x slang-2.3.0-7.fc25.s390x xz-libs-5.2.3-2.fc25.s390x zlib-1.2.8-10.fc25.s390x
  (gdb) where
  #0  0x0000000000000000 in ?? ()
  #1  0x00000000010fad6a in machines__deliver_event (machines=<optimized out>, machines@entry=0x2c6fd18,
      evlist=<optimized out>, event=event@entry=0x3fffdf00470, sample=0x3ffffffe880, sample@entry=0x3ffffffe888,
      tool=tool@entry=0x1312968 <build_id.mark_dso_hit_ops>, file_offset=1136) at util/session.c:1287
  #2  0x00000000010fbf4e in perf_session__deliver_event (file_offset=1136, tool=0x1312968 <build_id.mark_dso_hit_ops>,
      sample=0x3ffffffe888, event=0x3fffdf00470, session=0x2c6fc30) at util/session.c:1340
  #3  perf_session__process_event (session=0x2c6fc30, session@entry=0x0, event=event@entry=0x3fffdf00470,
      file_offset=file_offset@entry=1136) at util/session.c:1522
  #4  0x00000000010fddde in __perf_session__process_events (file_size=11880, data_size=<optimized out>,
      data_offset=<optimized out>, session=0x0) at util/session.c:1899
  #5  perf_session__process_events (session=0x0, session@entry=0x2c6fc30) at util/session.c:1953
  #6  0x000000000103b2ac in perf_session__list_build_ids (with_hits=<optimized out>, force=<optimized out>)
      at builtin-buildid-list.c:83
  #7  cmd_buildid_list (argc=<optimized out>, argv=<optimized out>) at builtin-buildid-list.c:115
  #8  0x00000000010a026c in run_builtin (p=0x1311f78 <commands+24>, argc=argc@entry=2, argv=argv@entry=0x3fffffff3c0)
      at perf.c:296
  #9  0x000000000102bc00 in handle_internal_command (argv=<optimized out>, argc=2) at perf.c:348
  #10 run_argv (argcp=<synthetic pointer>, argv=<synthetic pointer>) at perf.c:392
  #11 main (argc=<optimized out>, argv=0x3fffffff3c0) at perf.c:536
  (gdb)

Fix it by adding a stub event handler for namespace event.

Committer testing:

Further clarifying, plain using 'perf buildid-list' will not end up in a
SEGFAULT when processing a perf.data file with namespace info:

  # perf record -a --namespaces sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 2.024 MB perf.data (1058 samples) ]
  # perf buildid-list | wc -l
  38
  # perf buildid-list | head -5
  e2a171c7b905826fc8494f0711ba76ab6abbd604 /lib/modules/4.14.0-rc3+/build/vmlinux
  874840a02d8f8a31cedd605d0b8653145472ced3 /lib/modules/4.14.0-rc3+/kernel/arch/x86/kvm/kvm-intel.ko
  ea7223776730cd8a22f320040aae4d54312984bc /lib/modules/4.14.0-rc3+/kernel/drivers/gpu/drm/i915/i915.ko
  5961535e6732a8edb7f22b3f148bb2fa2e0be4b9 /lib/modules/4.14.0-rc3+/kernel/drivers/gpu/drm/drm.ko
  f045f54aa78cf1931cc893f78b6cbc52c72a8cb1 /usr/lib64/libc-2.25.so
  #

It is only when one asks for checking what of those entries actually had
samples, i.e. when we use either -H or --with-hits, that we will process
all the PERF_RECORD_ events, and since tools/perf/builtin-buildid-list.c
neither explicitely set a perf_tool.namespaces() callback nor the
default stub was set that we end up, when processing a
PERF_RECORD_NAMESPACE record, causing a SEGFAULT:

  # perf buildid-list -H
  Segmentation fault (core dumped)
  ^C
  #

Reported-and-Tested-by: Thomas-Mich Richter <tmricht@linux.vnet.ibm.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Hari Bathini <hbathini@linux.vnet.ibm.com>
Cc: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas-Mich Richter <tmricht@linux.vnet.ibm.com>
Fixes: f3b3614 ("perf tools: Add PERF_RECORD_NAMESPACES to include namespaces related info")
Link: http://lkml.kernel.org/r/20171017132900.11043-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

anholt pushed a commit that referenced this issue Feb 22, 2018

RDS: IB: Fix null pointer issue
Scenario:
1. Port down and do fail over
2. Ap do rds_bind syscall

PID: 47039  TASK: ffff89887e2fe640  CPU: 47  COMMAND: "kworker/u:6"
 #0 [ffff898e35f159f0] machine_kexec at ffffffff8103abf9
 #1 [ffff898e35f15a60] crash_kexec at ffffffff810b96e3
 #2 [ffff898e35f15b30] oops_end at ffffffff8150f518
 #3 [ffff898e35f15b60] no_context at ffffffff8104854c
 #4 [ffff898e35f15ba0] __bad_area_nosemaphore at ffffffff81048675
 #5 [ffff898e35f15bf0] bad_area_nosemaphore at ffffffff810487d3
 #6 [ffff898e35f15c00] do_page_fault at ffffffff815120b8
 #7 [ffff898e35f15d10] page_fault at ffffffff8150ea95
    [exception RIP: unknown or invalid address]
    RIP: 0000000000000000  RSP: ffff898e35f15dc8  RFLAGS: 00010282
    RAX: 00000000fffffffe  RBX: ffff889b77f6fc00  RCX:ffffffff81c99d88
    RDX: 0000000000000000  RSI: ffff896019ee08e8  RDI:ffff889b77f6fc00
    RBP: ffff898e35f15df0   R8: ffff896019ee08c8  R9:0000000000000000
    R10: 0000000000000400  R11: 0000000000000000  R12:ffff896019ee08c0
    R13: ffff889b77f6fe68  R14: ffffffff81c99d80  R15: ffffffffa022a1e0
    ORIG_RAX: ffffffffffffffff  CS: 0010 SS: 0018
 #8 [ffff898e35f15dc8] cma_ndev_work_handler at ffffffffa022a228 [rdma_cm]
 #9 [ffff898e35f15df8] process_one_work at ffffffff8108a7c6
 #10 [ffff898e35f15e58] worker_thread at ffffffff8108bda0
 #11 [ffff898e35f15ee8] kthread at ffffffff81090fe6

PID: 45659  TASK: ffff880d313d2500  CPU: 31  COMMAND: "oracle_45659_ap"
 #0 [ffff881024ccfc98] __schedule at ffffffff8150bac4
 #1 [ffff881024ccfd40] schedule at ffffffff8150c2cf
 #2 [ffff881024ccfd50] __mutex_lock_slowpath at ffffffff8150cee7
 #3 [ffff881024ccfdc0] mutex_lock at ffffffff8150cdeb
 #4 [ffff881024ccfde0] rdma_destroy_id at ffffffffa022a027 [rdma_cm]
 #5 [ffff881024ccfe10] rds_ib_laddr_check at ffffffffa0357857 [rds_rdma]
 #6 [ffff881024ccfe50] rds_trans_get_preferred at ffffffffa0324c2a [rds]
 #7 [ffff881024ccfe80] rds_bind at ffffffffa031d690 [rds]
 #8 [ffff881024ccfeb0] sys_bind at ffffffff8142a670

PID: 45659                          PID: 47039
rds_ib_laddr_check
  /* create id_priv with a null event_handler */
  rdma_create_id
  rdma_bind_addr
    cma_acquire_dev
      /* add id_priv to cma_dev->id_list */
      cma_attach_to_dev
                                    cma_ndev_work_handler
                                      /* event_hanlder is null */
                                      id_priv->id.event_handler

Signed-off-by: Guanglei Li <guanglei.li@oracle.com>
Signed-off-by: Honglei Wang <honglei.wang@oracle.com>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>
Reviewed-by: Yanjun Zhu <yanjun.zhu@oracle.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Acked-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

anholt pushed a commit that referenced this issue Mar 9, 2018

mm, swap, frontswap: fix THP swap if frontswap enabled
commit 7ba7166 upstream.

It was reported by Sergey Senozhatsky that if THP (Transparent Huge
Page) and frontswap (via zswap) are both enabled, when memory goes low
so that swap is triggered, segfault and memory corruption will occur in
random user space applications as follow,

kernel: urxvt[338]: segfault at 20 ip 00007fc08889ae0d sp 00007ffc73a7fc40 error 6 in libc-2.26.so[7fc08881a000+1ae000]
 #0  0x00007fc08889ae0d _int_malloc (libc.so.6)
 #1  0x00007fc08889c2f3 malloc (libc.so.6)
 #2  0x0000560e6004bff7 _Z14rxvt_wcstoutf8PKwi (urxvt)
 #3  0x0000560e6005e75c n/a (urxvt)
 #4  0x0000560e6007d9f1 _ZN16rxvt_perl_interp6invokeEP9rxvt_term9hook_typez (urxvt)
 #5  0x0000560e6003d988 _ZN9rxvt_term9cmd_parseEv (urxvt)
 #6  0x0000560e60042804 _ZN9rxvt_term6pty_cbERN2ev2ioEi (urxvt)
 #7  0x0000560e6005c10f _Z17ev_invoke_pendingv (urxvt)
 #8  0x0000560e6005cb55 ev_run (urxvt)
 #9  0x0000560e6003b9b9 main (urxvt)
 #10 0x00007fc08883af4a __libc_start_main (libc.so.6)
 #11 0x0000560e6003f9da _start (urxvt)

After bisection, it was found the first bad commit is bd4c82c ("mm,
THP, swap: delay splitting THP after swapped out").

The root cause is as follows:

When the pages are written to swap device during swapping out in
swap_writepage(), zswap (fontswap) is tried to compress the pages to
improve performance.  But zswap (frontswap) will treat THP as a normal
page, so only the head page is saved.  After swapping in, tail pages
will not be restored to their original contents, causing memory
corruption in the applications.

This is fixed by refusing to save page in the frontswap store functions
if the page is a THP.  So that the THP will be swapped out to swap
device.

Another choice is to split THP if frontswap is enabled.  But it is found
that the frontswap enabling isn't flexible.  For example, if
CONFIG_ZSWAP=y (cannot be module), frontswap will be enabled even if
zswap itself isn't enabled.

Frontswap has multiple backends, to make it easy for one backend to
enable THP support, the THP checking is put in backend frontswap store
functions instead of the general interfaces.

Link: http://lkml.kernel.org/r/20180209084947.22749-1-ying.huang@intel.com
Fixes: bd4c82c ("mm, THP, swap: delay splitting THP after swapped out")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Tested-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Suggested-by: Minchan Kim <minchan@kernel.org>	[put THP checking in backend]
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Shaohua Li <shli@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: <stable@vger.kernel.org>	[4.14]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

stschake pushed a commit to stschake/linux that referenced this issue Mar 11, 2018

mm, swap, frontswap: fix THP swap if frontswap enabled
It was reported by Sergey Senozhatsky that if THP (Transparent Huge
Page) and frontswap (via zswap) are both enabled, when memory goes low
so that swap is triggered, segfault and memory corruption will occur in
random user space applications as follow,

kernel: urxvt[338]: segfault at 20 ip 00007fc08889ae0d sp 00007ffc73a7fc40 error 6 in libc-2.26.so[7fc08881a000+1ae000]
 #0  0x00007fc08889ae0d _int_malloc (libc.so.6)
 anholt#1  0x00007fc08889c2f3 malloc (libc.so.6)
 anholt#2  0x0000560e6004bff7 _Z14rxvt_wcstoutf8PKwi (urxvt)
 anholt#3  0x0000560e6005e75c n/a (urxvt)
 anholt#4  0x0000560e6007d9f1 _ZN16rxvt_perl_interp6invokeEP9rxvt_term9hook_typez (urxvt)
 anholt#5  0x0000560e6003d988 _ZN9rxvt_term9cmd_parseEv (urxvt)
 anholt#6  0x0000560e60042804 _ZN9rxvt_term6pty_cbERN2ev2ioEi (urxvt)
 anholt#7  0x0000560e6005c10f _Z17ev_invoke_pendingv (urxvt)
 anholt#8  0x0000560e6005cb55 ev_run (urxvt)
 anholt#9  0x0000560e6003b9b9 main (urxvt)
 anholt#10 0x00007fc08883af4a __libc_start_main (libc.so.6)
 anholt#11 0x0000560e6003f9da _start (urxvt)

After bisection, it was found the first bad commit is bd4c82c ("mm,
THP, swap: delay splitting THP after swapped out").

The root cause is as follows:

When the pages are written to swap device during swapping out in
swap_writepage(), zswap (fontswap) is tried to compress the pages to
improve performance.  But zswap (frontswap) will treat THP as a normal
page, so only the head page is saved.  After swapping in, tail pages
will not be restored to their original contents, causing memory
corruption in the applications.

This is fixed by refusing to save page in the frontswap store functions
if the page is a THP.  So that the THP will be swapped out to swap
device.

Another choice is to split THP if frontswap is enabled.  But it is found
that the frontswap enabling isn't flexible.  For example, if
CONFIG_ZSWAP=y (cannot be module), frontswap will be enabled even if
zswap itself isn't enabled.

Frontswap has multiple backends, to make it easy for one backend to
enable THP support, the THP checking is put in backend frontswap store
functions instead of the general interfaces.

Link: http://lkml.kernel.org/r/20180209084947.22749-1-ying.huang@intel.com
Fixes: bd4c82c ("mm, THP, swap: delay splitting THP after swapped out")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Tested-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Suggested-by: Minchan Kim <minchan@kernel.org>	[put THP checking in backend]
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Shaohua Li <shli@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: <stable@vger.kernel.org>	[4.14]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

stschake pushed a commit to stschake/linux that referenced this issue Mar 16, 2018

signals: Move put_compat_sigset to compat.h to silence hardened usercopy
Since commit afcc90f ("usercopy: WARN() on slab cache usercopy
region violations"), MIPS systems booting with a compat root filesystem
emit a warning when copying compat siginfo to userspace:

WARNING: CPU: 0 PID: 953 at mm/usercopy.c:81 usercopy_warn+0x98/0xe8
Bad or missing usercopy whitelist? Kernel memory exposure attempt
detected from SLAB object 'task_struct' (offset 1432, size 16)!
Modules linked in:
CPU: 0 PID: 953 Comm: S01logging Not tainted 4.16.0-rc2 anholt#10
Stack : ffffffff808c0000 0000000000000000 0000000000000001 65ac85163f3bdc4a
	65ac85163f3bdc4a 0000000000000000 90000000ff667ab8 ffffffff808c0000
	00000000000003f8 ffffffff808d0000 00000000000000d1 0000000000000000
	000000000000003c 0000000000000000 ffffffff808c8ca8 ffffffff808d0000
	ffffffff808d0000 ffffffff80810000 fffffc0000000000 ffffffff80785c30
	0000000000000009 0000000000000051 90000000ff667eb0 90000000ff667db0
	000000007fe0d938 0000000000000018 ffffffff80449958 0000000020052798
	ffffffff808c0000 90000000ff664000 90000000ff667ab0 00000000100c0000
	ffffffff80698810 0000000000000000 0000000000000000 0000000000000000
	0000000000000000 0000000000000000 ffffffff8010d02c 65ac85163f3bdc4a
	...
Call Trace:
[<ffffffff8010d02c>] show_stack+0x9c/0x130
[<ffffffff80698810>] dump_stack+0x90/0xd0
[<ffffffff80137b78>] __warn+0x100/0x118
[<ffffffff80137bdc>] warn_slowpath_fmt+0x4c/0x70
[<ffffffff8021e4a8>] usercopy_warn+0x98/0xe8
[<ffffffff8021e68c>] __check_object_size+0xfc/0x250
[<ffffffff801bbfb8>] put_compat_sigset+0x30/0x88
[<ffffffff8011af24>] setup_rt_frame_n32+0xc4/0x160
[<ffffffff8010b8b4>] do_signal+0x19c/0x230
[<ffffffff8010c408>] do_notify_resume+0x60/0x78
[<ffffffff80106f50>] work_notifysig+0x10/0x18
---[ end trace 88fffbf69147f48a ]---

Commit 5905429 ("fork: Provide usercopy whitelisting for
task_struct") noted that:

"While the blocked and saved_sigmask fields of task_struct are copied to
userspace (via sigmask_to_save() and setup_rt_frame()), it is always
copied with a static length (i.e. sizeof(sigset_t))."

However, this is not true in the case of compat signals, whose sigset
is copied by put_compat_sigset and receives size as an argument.

At most call sites, put_compat_sigset is copying a sigset from the
current task_struct. This triggers a warning when
CONFIG_HARDENED_USERCOPY is active. However, by marking this function as
static inline, the warning can be avoided because in all of these cases
the size is constant at compile time, which is allowed. The only site
where this is not the case is handling the rt_sigpending syscall, but
there the copy is being made from a stack local variable so does not
trigger the warning.

Move put_compat_sigset to compat.h, and mark it static inline. This
fixes the WARN on MIPS.

Fixes: afcc90f ("usercopy: WARN() on slab cache usercopy region violations")
Signed-off-by: Matt Redfearn <matt.redfearn@mips.com>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: "Dmitry V . Levin" <ldv@altlinux.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: kernel-hardening@lists.openwall.com
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/18639/
Signed-off-by: James Hogan <jhogan@kernel.org>

anholt pushed a commit that referenced this issue Apr 23, 2018

include/linux/kexec.h: silence compile warnings
Patch series "kexec_file: Clean up purgatory load", v2.

Following the discussion with Dave and AKASHI, here are the common code
patches extracted from my recent patch set (Add kexec_file_load support
to s390) [1].  The patches were extracted to allow upstream integration
together with AKASHI's common code patches before the arch code gets
adjusted to the new base.

The reason for this series is to prepare common code for adding
kexec_file_load to s390 as well as cleaning up the mis-use of the
sh_offset field during purgatory load.  In detail this series contains:

Patch #1&2: Minor cleanups/fixes.

Patch #3-9: Clean up the purgatory load/relocation code.  Especially
remove the mis-use of the purgatory_info->sechdrs->sh_offset field,
currently holding a pointer into either kexec_purgatory (ro) or
purgatory_buf (rw) depending on the section.  With these patches the
section address will be calculated verbosely and sh_offset will contain
the offset of the section in the stripped purgatory binary
(purgatory_buf).

Patch #10: Allows architectures to set the purgatory load address.  This
patch is important for s390 as the kernel and purgatory have to be
loaded to fixed addresses.  In current code this is impossible as the
purgatory load is opaque to the architecture.

Patch #11: Moves x86 purgatories sha implementation to common lib/
directory to allow reuse in other architectures.

This patch (of 11)

When building the kernel with CONFIG_KEXEC_FILE enabled gcc prints a
compile warning multiple times.

  In file included from <path>/linux/init/initramfs.c:526:0:
  <path>/include/linux/kexec.h:120:9: warning: `struct kimage' declared inside parameter list [enabled by default]
           unsigned long cmdline_len);
           ^

This is because the typedefs for kexec_file_load uses struct kimage
before it is declared.  Fix this by simply forward declaring struct
kimage.

Link: http://lkml.kernel.org/r/20180321112751.22196-2-prudo@linux.vnet.ibm.com
Signed-off-by: Philipp Rudo <prudo@linux.vnet.ibm.com>
Acked-by: Dave Young <dyoung@redhat.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Thiago Jung Bauermann <bauerman@linux.vnet.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

anholt pushed a commit that referenced this issue May 7, 2018

RDS: IB: Fix null pointer issue
[ Upstream commit 2c0aa08 ]

Scenario:
1. Port down and do fail over
2. Ap do rds_bind syscall

PID: 47039  TASK: ffff89887e2fe640  CPU: 47  COMMAND: "kworker/u:6"
 #0 [ffff898e35f159f0] machine_kexec at ffffffff8103abf9
 #1 [ffff898e35f15a60] crash_kexec at ffffffff810b96e3
 #2 [ffff898e35f15b30] oops_end at ffffffff8150f518
 #3 [ffff898e35f15b60] no_context at ffffffff8104854c
 #4 [ffff898e35f15ba0] __bad_area_nosemaphore at ffffffff81048675
 #5 [ffff898e35f15bf0] bad_area_nosemaphore at ffffffff810487d3
 #6 [ffff898e35f15c00] do_page_fault at ffffffff815120b8
 #7 [ffff898e35f15d10] page_fault at ffffffff8150ea95
    [exception RIP: unknown or invalid address]
    RIP: 0000000000000000  RSP: ffff898e35f15dc8  RFLAGS: 00010282
    RAX: 00000000fffffffe  RBX: ffff889b77f6fc00  RCX:ffffffff81c99d88
    RDX: 0000000000000000  RSI: ffff896019ee08e8  RDI:ffff889b77f6fc00
    RBP: ffff898e35f15df0   R8: ffff896019ee08c8  R9:0000000000000000
    R10: 0000000000000400  R11: 0000000000000000  R12:ffff896019ee08c0
    R13: ffff889b77f6fe68  R14: ffffffff81c99d80  R15: ffffffffa022a1e0
    ORIG_RAX: ffffffffffffffff  CS: 0010 SS: 0018
 #8 [ffff898e35f15dc8] cma_ndev_work_handler at ffffffffa022a228 [rdma_cm]
 #9 [ffff898e35f15df8] process_one_work at ffffffff8108a7c6
 #10 [ffff898e35f15e58] worker_thread at ffffffff8108bda0
 #11 [ffff898e35f15ee8] kthread at ffffffff81090fe6

PID: 45659  TASK: ffff880d313d2500  CPU: 31  COMMAND: "oracle_45659_ap"
 #0 [ffff881024ccfc98] __schedule at ffffffff8150bac4
 #1 [ffff881024ccfd40] schedule at ffffffff8150c2cf
 #2 [ffff881024ccfd50] __mutex_lock_slowpath at ffffffff8150cee7
 #3 [ffff881024ccfdc0] mutex_lock at ffffffff8150cdeb
 #4 [ffff881024ccfde0] rdma_destroy_id at ffffffffa022a027 [rdma_cm]
 #5 [ffff881024ccfe10] rds_ib_laddr_check at ffffffffa0357857 [rds_rdma]
 #6 [ffff881024ccfe50] rds_trans_get_preferred at ffffffffa0324c2a [rds]
 #7 [ffff881024ccfe80] rds_bind at ffffffffa031d690 [rds]
 #8 [ffff881024ccfeb0] sys_bind at ffffffff8142a670

PID: 45659                          PID: 47039
rds_ib_laddr_check
  /* create id_priv with a null event_handler */
  rdma_create_id
  rdma_bind_addr
    cma_acquire_dev
      /* add id_priv to cma_dev->id_list */
      cma_attach_to_dev
                                    cma_ndev_work_handler
                                      /* event_hanlder is null */
                                      id_priv->id.event_handler

Signed-off-by: Guanglei Li <guanglei.li@oracle.com>
Signed-off-by: Honglei Wang <honglei.wang@oracle.com>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>
Reviewed-by: Yanjun Zhu <yanjun.zhu@oracle.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Acked-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

anholt pushed a commit that referenced this issue Jun 29, 2018

Shirish S Alex Deucher
drm/amdgpu: avoid sleep while executing atombios table (V2)
This patch replaces kzalloc's flag from GFP_KERNEL to
GFP_ATOMIC to avoid sleeping in atomic context.

Below is the stack trace:

BUG: sleeping function called from invalid context at mm/slab.h:***
in_atomic(): 1, irqs_disabled(): 0, pid: 1137, name: DrmThread
CPU: 1 PID: 1137 Comm: DrmThread Tainted: G        W       4.14.43 #10
Call Trace:
 dump_stack+0x4d/0x63
 ___might_sleep+0x11f/0x12e
 __kmalloc+0x76/0x126
 amdgpu_atom_execute_table_locked+0xfc/0x285
 amdgpu_atom_execute_table+0x5d/0x72
 transmitter_control_v1_5+0xef/0x11a
 hwss_edp_backlight_control+0x132/0x151
 dce110_disable_stream+0x133/0x16e
 core_link_disable_stream+0x1c5/0x23b
 dce110_reset_hw_ctx_wrap+0xb4/0x1aa
 dce110_apply_ctx_to_hw+0x4e/0x6da
 ? generic_reg_get+0x1f/0x33
 dc_commit_state+0x33f/0x3d2
 amdgpu_dm_atomic_commit_tail+0x2cf/0x5d2
 ? wait_for_common+0x5b/0x69
 commit_tail+0x42/0x64
 drm_atomic_helper_commit+0xdc/0xf9
 drm_atomic_helper_set_config+0x5c/0x76
 __drm_mode_set_config_internal+0x64/0x105
 drm_mode_setcrtc+0x474/0x56f
 ? drm_mode_getcrtc+0x155/0x155
 drm_ioctl_kernel+0x6c/0xa8
 drm_ioctl+0x267/0x353
 ? drm_mode_getcrtc+0x155/0x155
 amdgpu_drm_ioctl+0x4f/0x7f
 vfs_ioctl+0x21/0x2f
 do_vfs_ioctl+0x4c4/0x4e7
 ? security_file_ioctl+0x3b/0x4f
 SyS_ioctl+0x57/0x79
 do_syscall_64+0x64/0x72
 entry_SYSCALL_64_after_hwframe+0x3d/0xa2

V2: Added stack trace in commit message.

Signed-off-by: Shirish S <shirish.s@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
@anholt

This comment has been minimized.

Owner

anholt commented Oct 30, 2018

Progress on this: hx8357d tinydrm driver merged, kmsro patch series submitted.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment