diff mbox series

[net] net/sched: sch_netem: Fix arithmetic in netem_dump() for 32-bit platforms

Message ID 20220616234336.2443-1-yepeilin.cs@gmail.com (mailing list archive)
State Accepted
Commit a2b1a5d40bd12b44322c2ccd40bb0ec1699708b6
Delegated to: Netdev Maintainers
Headers show
Series [net] net/sched: sch_netem: Fix arithmetic in netem_dump() for 32-bit platforms | expand

Checks

Context Check Description
netdev/tree_selection success Clearly marked for net
netdev/fixes_present success Fixes tag present in non-next series
netdev/subject_prefix success Link
netdev/cover_letter success Single patches do not need cover letters
netdev/patch_count success Link
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 0 this patch: 0
netdev/cc_maintainers success CCed 10 of 10 maintainers
netdev/build_clang success Errors and warnings before: 0 this patch: 0
netdev/module_param success Was 0 now: 0
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success Fixes tag looks correct
netdev/build_allmodconfig_warn success Errors and warnings before: 0 this patch: 0
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 11 lines checked
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

Peilin Ye June 16, 2022, 11:43 p.m. UTC
From: Peilin Ye <peilin.ye@bytedance.com>

As reported by Yuming, currently tc always show a latency of UINT_MAX
for netem Qdisc's on 32-bit platforms:

    $ tc qdisc add dev dummy0 root netem latency 100ms
    $ tc qdisc show dev dummy0
    qdisc netem 8001: root refcnt 2 limit 1000 delay 275s  275s
                                               ^^^^^^^^^^^^^^^^

Let us take a closer look at netem_dump():

        qopt.latency = min_t(psched_tdiff_t, PSCHED_NS2TICKS(q->latency,
                             UINT_MAX);

qopt.latency is __u32, psched_tdiff_t is signed long,
(psched_tdiff_t)(UINT_MAX) is negative for 32-bit platforms, so
qopt.latency is always UINT_MAX.

Fix it by using psched_time_t (u64) instead.

Note: confusingly, users have two ways to specify 'latency':

  1. normally, via '__u32 latency' in struct tc_netem_qopt;
  2. via the TCA_NETEM_LATENCY64 attribute, which is s64.

For the second case, theoretically 'latency' could be negative.  This
patch ignores that corner case, since it is broken (i.e. assigning a
negative s64 to __u32) anyways, and should be handled separately.

Thanks Ted Lin for the analysis [1] .

[1] https://github.com/raspberrypi/linux/issues/3512

Reported-by: Yuming Chen <chenyuming.junnan@bytedance.com>
Fixes: 112f9cb65643 ("netem: convert to qdisc_watchdog_schedule_ns")
Reviewed-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
---
 net/sched/sch_netem.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Stephen Hemminger June 17, 2022, 3:08 a.m. UTC | #1
On Thu, 16 Jun 2022 16:43:36 -0700
Peilin Ye <yepeilin.cs@gmail.com> wrote:

> From: Peilin Ye <peilin.ye@bytedance.com>
> 
> As reported by Yuming, currently tc always show a latency of UINT_MAX
> for netem Qdisc's on 32-bit platforms:
> 
>     $ tc qdisc add dev dummy0 root netem latency 100ms
>     $ tc qdisc show dev dummy0
>     qdisc netem 8001: root refcnt 2 limit 1000 delay 275s  275s
>                                                ^^^^^^^^^^^^^^^^
> 
> Let us take a closer look at netem_dump():
> 
>         qopt.latency = min_t(psched_tdiff_t, PSCHED_NS2TICKS(q->latency,
>                              UINT_MAX);
> 
> qopt.latency is __u32, psched_tdiff_t is signed long,
> (psched_tdiff_t)(UINT_MAX) is negative for 32-bit platforms, so
> qopt.latency is always UINT_MAX.
> 
> Fix it by using psched_time_t (u64) instead.
> 
> Note: confusingly, users have two ways to specify 'latency':
> 
>   1. normally, via '__u32 latency' in struct tc_netem_qopt;
>   2. via the TCA_NETEM_LATENCY64 attribute, which is s64.
> 
> For the second case, theoretically 'latency' could be negative.  This
> patch ignores that corner case, since it is broken (i.e. assigning a
> negative s64 to __u32) anyways, and should be handled separately.
> 
> Thanks Ted Lin for the analysis [1] .
> 
> [1] https://github.com/raspberrypi/linux/issues/3512
> 
> Reported-by: Yuming Chen <chenyuming.junnan@bytedance.com>
> Fixes: 112f9cb65643 ("netem: convert to qdisc_watchdog_schedule_ns")
> Reviewed-by: Cong Wang <cong.wang@bytedance.com>
> Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
> ---
>  net/sched/sch_netem.c | 4 ++--

Thanks for fixing. 
Guess it is time to run netem on one of the Pi's.

Acked-by: Stephen Hemminger <stephen@networkplumber.org>
patchwork-bot+netdevbpf@kernel.org June 18, 2022, 3:40 a.m. UTC | #2
Hello:

This patch was applied to netdev/net.git (master)
by Jakub Kicinski <kuba@kernel.org>:

On Thu, 16 Jun 2022 16:43:36 -0700 you wrote:
> From: Peilin Ye <peilin.ye@bytedance.com>
> 
> As reported by Yuming, currently tc always show a latency of UINT_MAX
> for netem Qdisc's on 32-bit platforms:
> 
>     $ tc qdisc add dev dummy0 root netem latency 100ms
>     $ tc qdisc show dev dummy0
>     qdisc netem 8001: root refcnt 2 limit 1000 delay 275s  275s
>                                                ^^^^^^^^^^^^^^^^
> 
> [...]

Here is the summary with links:
  - [net] net/sched: sch_netem: Fix arithmetic in netem_dump() for 32-bit platforms
    https://git.kernel.org/netdev/net/c/a2b1a5d40bd1

You are awesome, thank you!
diff mbox series

Patch

diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index ed4ccef5d6a8..5449ed114e40 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -1146,9 +1146,9 @@  static int netem_dump(struct Qdisc *sch, struct sk_buff *skb)
 	struct tc_netem_rate rate;
 	struct tc_netem_slot slot;
 
-	qopt.latency = min_t(psched_tdiff_t, PSCHED_NS2TICKS(q->latency),
+	qopt.latency = min_t(psched_time_t, PSCHED_NS2TICKS(q->latency),
 			     UINT_MAX);
-	qopt.jitter = min_t(psched_tdiff_t, PSCHED_NS2TICKS(q->jitter),
+	qopt.jitter = min_t(psched_time_t, PSCHED_NS2TICKS(q->jitter),
 			    UINT_MAX);
 	qopt.limit = q->limit;
 	qopt.loss = q->loss;