socket: fix option SO_TIMESTAMPING_NEW
Subsystems (3)
PATCH v2 1/2 socket: fix option SO_TIMESTAMPING_NEW
The comparison of optname with SO_TIMESTAMPING_NEW is wrong way around, so SOCK_TSTAMP_NEW will first be set and than reset again. Additionally move it out of the test for SOF_TIMESTAMPING_RX_SOFTWARE as this seems unrelated.
This problem happens on 32 bit platforms were the libc has already switched to struct timespec64 (from SO_TIMExxx_OLD to SO_TIMExxx_NEW socket options). ptp4l complains with "missing timestamp on transmitted peer delay request" because the wrong format is received (and discarded).
---
v2:
-----
- integrated proposal from Willem de Bruijn
- added Reviewed-by: from Willem and Deepa
Quoted 2 lines · Willem de Bruijn
This suggested fix still sets and clears the flag if calling
SO_TIMESTAMPING_NEWto disable timestamping.
where is it cleared?
Review
Quoted 10 lines
Instead, how about
case
SO_TIMESTAMPING_NEW:(quoted diff) +3
-1- sock_set_flag(sk, SOCK_TSTAMP_NEW); fallthrough; case SO_TIMESTAMPING_OLD: + sock_valbool_flag(sk, SOCK_TSTAMP_NEW, + optname == SO_TIMESTAMPING_NEW); +
using you version looks clearer
Quoted 2 lines
if (val & SOF_TIMESTAMPING_OPT_ID &&
I would like to keep this below the "ret = -FOO; break" statements. IMHO the setsockopt() call should either completely fail or succeed.
net/core/sock.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-)
net/core/sock.c +3 -7
@@ -994,8 +994,6 @@ int sock_setsockopt(struct socket *sock, int level, int optname,__sock_set_timestamps (sk, valbool, true, true);break ;case SO_TIMESTAMPING_NEW : -sock_set_flag (sk,SOCK_TSTAMP_NEW ); - fallthrough;case SO_TIMESTAMPING_OLD :if (val & ~SOF_TIMESTAMPING_MASK ) { ret = -EINVAL ;
@@ -1024,16 +1022,14 @@ int sock_setsockopt(struct socket *sock, int level, int optname, } sk->sk_tsflags = val; +sock_valbool_flag (sk,SOCK_TSTAMP_NEW , optname ==SO_TIMESTAMPING_NEW ); +if (val &SOF_TIMESTAMPING_RX_SOFTWARE )sock_enable_timestamp (sk,SOCK_TIMESTAMPING_RX_SOFTWARE ); -else { -if (optname ==SO_TIMESTAMPING_NEW ) -sock_reset_flag (sk,SOCK_TSTAMP_NEW ); - +else sock_disable_timestamp (sk, (1UL <<SOCK_TIMESTAMPING_RX_SOFTWARE )); - }break ;case SO_RCVLOWAT :
Signature · 8 lines
Christian Eggers Embedded software developer Arnold & Richter Cine Technik GmbH & Co. Betriebs KG Sitz: Muenchen - Registergericht: Amtsgericht Muenchen - Handelsregisternummer: HRA 57918 Persoenlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH Sitz: Muenchen - Registergericht: Amtsgericht Muenchen - Handelsregisternummer: HRB 54477 Geschaeftsfuehrer: Dr. Michael Neuhaeuser; Stephan Schenk; Walter Trauninger; Markus Zeiler
PATCH v2 2/2 socket: don't clear SOCK_TSTAMP_NEW when SO_TIMESTAMPNS is disabled
SOCK_TSTAMP_NEW (timespec64 instead of timespec) is also used for hardware time stamps (configured via SO_TIMESTAMPING_NEW).
User space (ptp4l) first configures hardware time stamping via SO_TIMESTAMPING_NEW which sets SOCK_TSTAMP_NEW. In the next step, ptp4l disables SO_TIMESTAMPNS(_NEW) (software time stamps), but this must not switch hardware time stamps back to "32 bit mode".
This problem happens on 32 bit platforms were the libc has already switched to struct timespec64 (from SO_TIMExxx_OLD to SO_TIMExxx_NEW socket options). ptp4l complains with "missing timestamp on transmitted peer delay request" because the wrong format is received (and discarded).
---
v2:
-----
- Added ACKs from Willem and Deepa
Quoted 11 lines · Deepa Dinamani
Quoted 3 lines · Willem de Bruijn
As this commit message shows, with
SO_TIMESTAMP(NS)andSO_TIMESTAMPINGthat can be independently turned on and off, disabling one can incorrectly switch modes while the other is still active.This will not help the case when a child process that inherits the fd and then sets SO_TIMESTAMP*_OLD/NEW on it, while the parent uses the other version.
One of the two processes might still be surprised. But, child and parent actively using the same socket fd might be expecting trouble already.
Usually the decision between SO_TIMESTAMP*_OLD/NEW will be done at compile time (trough the system C library headers). For a typical "distribution" it should be quite unlikely that two programs will be compiled using different settings.
net/core/sock.c | 1 - 1 file changed, 1 deletion(-)
net/core/sock.c +0 -1
@@ -757,7 +757,6 @@ static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns) }else {sock_reset_flag (sk,SOCK_RCVTSTAMP );sock_reset_flag (sk,SOCK_RCVTSTAMPNS ); -sock_reset_flag (sk,SOCK_TSTAMP_NEW ); } }
Signature · 8 lines
Christian Eggers Embedded software developer Arnold & Richter Cine Technik GmbH & Co. Betriebs KG Sitz: Muenchen - Registergericht: Amtsgericht Muenchen - Handelsregisternummer: HRA 57918 Persoenlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH Sitz: Muenchen - Registergericht: Amtsgericht Muenchen - Handelsregisternummer: HRB 54477 Geschaeftsfuehrer: Dr. Michael Neuhaeuser; Stephan Schenk; Walter Trauninger; Markus Zeiler
Quoted 12 lines · Christian Eggers
The comparison of optname with
SO_TIMESTAMPING_NEWis wrong way around, soSOCK_TSTAMP_NEWwill first be set and than reset again. Additionally move it out of the test forSOF_TIMESTAMPING_RX_SOFTWAREas this seems unrelated.This problem happens on 32 bit platforms were the libc has already switched to struct timespec64 (from
SO_TIMExxx_OLDtoSO_TIMExxx_NEWsocket options). ptp4l complains with "missing timestamp on transmitted peer delay request" because the wrong format is received (and discarded).
Quoted 4 lines
Fixes 9718475e6908 ("socket: Add SO_TIMESTAMPING_NEW")Signed-off-by Christian Eggers <[email protected]>Reviewed-by Willem de Bruijn <[email protected]>Reviewed-by Deepa Dinamani <[email protected]>
We have not yet reviewed this second patch. Please do not add such tags on behalf of other people.
That said, I now have and do agree with the change, so
Quoted 22 lines
---
v2:
-----
- integrated proposal from Willem de Bruijn
- added Reviewed-by: from Willem and DeepaQuoted 2 lines · Willem de Bruijn
This suggested fix still sets and clears the flag if calling
SO_TIMESTAMPING_NEWto disable timestamping.where is it cleared?
Review
Quoted 10 lines
Instead, how about
case
SO_TIMESTAMPING_NEW:(quoted diff) +3
-1- sock_set_flag(sk, SOCK_TSTAMP_NEW); fallthrough; case SO_TIMESTAMPING_OLD: + sock_valbool_flag(sk, SOCK_TSTAMP_NEW, + optname == SO_TIMESTAMPING_NEW); +
Quoted 6 lines
using you version looks clearer
Quoted 2 lines
if (val & SOF_TIMESTAMPING_OPT_ID &&I would like to keep this below the "ret = -FOO; break" statements. IMHO the
setsockopt()call should either completely fail or succeed.
Agreed.
Quoted 46 lines
net/core/sock.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-)net/core/sock.c +3
-8@@ -994,8 +994,6 @@ int sock_setsockopt(struct socket *sock, int level, int optname,__sock_set_timestamps (sk, valbool, true, true);break ;case SO_TIMESTAMPING_NEW : -sock_set_flag (sk,SOCK_TSTAMP_NEW ); - fallthrough;case SO_TIMESTAMPING_OLD :if (val & ~SOF_TIMESTAMPING_MASK ) { ret = -EINVAL ;@@ -1024,16 +1022,14 @@ int sock_setsockopt(struct socket *sock, int level, int optname, } sk->sk_tsflags = val; +sock_valbool_flag (sk,SOCK_TSTAMP_NEW , optname ==SO_TIMESTAMPING_NEW ); +if (val &SOF_TIMESTAMPING_RX_SOFTWARE )sock_enable_timestamp (sk,SOCK_TIMESTAMPING_RX_SOFTWARE ); -else { -if (optname ==SO_TIMESTAMPING_NEW ) -sock_reset_flag (sk,SOCK_TSTAMP_NEW ); - +else sock_disable_timestamp (sk, (1UL <<SOCK_TIMESTAMPING_RX_SOFTWARE )); - }break ;case SO_RCVLOWAT : --Christian Eggers
Embedded software developerArnold & Richter Cine Technik GmbH & Co. Betriebs KG
Sitz: Muenchen - Registergericht: Amtsgericht Muenchen - Handelsregisternummer: HRA 57918 Persoenlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH
Sitz: Muenchen - Registergericht: Amtsgericht Muenchen - Handelsregisternummer: HRB 54477
Geschaeftsfuehrer: Dr. Michael Neuhaeuser; Stephan Schenk; Walter Trauninger; Markus Zeiler
Deepa Dinamani
Acked-byQuoted 5 lines · Christian Eggers
Quoted 4 lines
v2:
-----
- integrated proposal from Willem de Bruijn
- added Reviewed-by: from Willem and Deepa
You may add my
-Deepa
Quoted 8 lines · On Mon, 12 Oct 2020 22:10:31 -0700 Deepa Dinamani wrote:
Quoted 5 lines · Christian Eggers
Quoted 4 lines
v2:
-----
- integrated proposal from Willem de Bruijn
- added Reviewed-by: from Willem and DeepaYou may add my
Acked-by Deepa Dinamani <[email protected]>
Applied both, thanks everyone!