musl: Update to latest trunk

PATCH In review Khem Raj 3 participants 5 messages Nov 29, 2018 – Jan 21, 2019

Share & tools

Khem Raj

author

2018-11-29 17:52 UTC #0 lore raw

PATCH v2 1/1 musl: Update to latest trunk

Complete changelogs are here
https://git.musl-libc.org/cgit/musl/log/?qt=range&q=c50985d5c8e316c5c464f352e79eeebfed1121a9..39ef612aa193cc6e954ac5a01574300ccd4b7ef9

Signed-off-by Khem Raj <[email protected]>

---
V2: move to latest master ( 3 more commits for regressions )

 meta/recipes-core/musl/musl_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

meta/recipes-core/musl/musl_git.bb +1 -1

index 0d8f8eb2a4..b416ec45bf 100644
@@ -4,7 +4,7 @@
 require musl.inc
 inherit linuxloader
 
-SRCREV = "c50985d5c8e316c5c464f352e79eeebfed1121a9"
+SRCREV = "39ef612aa193cc6e954ac5a01574300ccd4b7ef9"
 
 PV = "1.1.20+git${SRCPV}"

Signature · 1 line

2.19.2

Richard Purdie

2018-11-30 17:04 UTC #1 lore raw

Quoted 6 lines · Khem Raj

Complete changelogs are here
https://git.musl-libc.org/cgit/musl/log/?qt=range&q=c50985d5c8e316c5c464f352e79eeebfed1121a9..39ef612aa193cc6e954ac5a01574300ccd4b7ef9
Signed-off-by Khem Raj <[email protected]>
---
V2: move to latest master ( 3 more commits for regressions )

Sorry, still fails :( :

https://autobuilder.yoctoproject.org/typhoon/#/builders/64/builds/39
https://autobuilder.yoctoproject.org/typhoon/#/builders/45/builds/39

Cheers,

Richard

André Draszik

2019-01-21 09:14 UTC #2 lore raw

In reply to Khem Raj

Hi,

Quoted 16 lines · Khem Raj

Complete changelogs are here
https://git.musl-libc.org/cgit/musl/log/?qt=range&q=c50985d5c8e316c5c464f352e79eeebfed1121a9..39ef612aa193cc6e954ac5a01574300ccd4b7ef9
Signed-off-by Khem Raj <[email protected]>
---
V2: move to latest master ( 3 more commits for regressions )
 meta/recipes-core/musl/musl_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/recipes- renamed from meta/recipes-core/musl/musl_git.bb +0 -0
core/musl/musl_git.bb
index 0d8f8eb2a4..b416ec45bf 100644

meta/recipes-core/musl/musl_git.bb line 7

Quoted 5 lines

meta/recipes-core/musl/musl_git.bb +1 -1

@@ -4,7 +4,7 @@
 require musl.inc
 inherit linuxloader
 
-SRCREV = "c50985d5c8e316c5c464f352e79eeebfed1121a9"
+SRCREV = "39ef612aa193cc6e954ac5a01574300ccd4b7ef9"

With this, link detection doesn't seem to work correctly anymore, instead constant messages in the syslog:

Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver [Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready
fec 30bf0000.ethernet eth1: Link is Up - 1Gbps/Full - flow control rx/tx
IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver [Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready
fec 30bf0000.ethernet eth1: Link is Up - 1Gbps/Full - flow control rx/tx
IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver [Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)

etc.

I've tracked it down to commit d1395c43c019 ("allow freeaddrinfo of arbitrary sublists of addrinfo list" in musl.

Even latest musl master de7dc1318f49 doesn't work for me with the same problem. The only way to make it work is to revert d1395c43c019 in musl, even with latest master.

link detection should be a kernel thing, or not? So how can a user-space change affect the kernel's link detection?

Any ideas?

Cheers,
Andre'

Khem Raj

author

2019-01-21 16:17 UTC #3 lore raw

Hi André

this feature was implemented but does not emulate full glibc behavior and its possible that app is expecting the glibc behavior, I have fixed nfsutils app in past see for details

https://patchwork.kernel.org/patch/10707759/

I think you can start with suspecting app, I dont rule out libc bug but thats less likely

Quoted 58 lines · André Draszik

Hi,
Quoted 16 lines · Khem Raj
Complete changelogs are here
https://git.musl-libc.org/cgit/musl/log/?qt=range&q=c50985d5c8e316c5c464f352e79eeebfed1121a9..39ef612aa193cc6e954ac5a01574300ccd4b7ef9
Signed-off-by Khem Raj <[email protected]>
---
V2: move to latest master ( 3 more commits for regressions )
 meta/recipes-core/musl/musl_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/recipes- renamed from meta/recipes-core/musl/musl_git.bb +0 -0
core/musl/musl_git.bb
index 0d8f8eb2a4..b416ec45bf 100644
meta/recipes-core/musl/musl_git.bb line 7
Quoted 5 lines
meta/recipes-core/musl/musl_git.bb +1 -1
@@ -4,7 +4,7 @@
 require musl.inc
 inherit linuxloader
 
-SRCREV = "c50985d5c8e316c5c464f352e79eeebfed1121a9"
+SRCREV = "39ef612aa193cc6e954ac5a01574300ccd4b7ef9"
With this, link detection doesn't seem to work correctly anymore, instead constant messages in the syslog:
Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver [Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready
fec 30bf0000.ethernet eth1: Link is Up - 1Gbps/Full - flow control rx/tx
IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver [Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready
fec 30bf0000.ethernet eth1: Link is Up - 1Gbps/Full - flow control rx/tx
IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver [Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
etc.
I've tracked it down to commit d1395c43c019 ("allow freeaddrinfo of arbitrary sublists of addrinfo list" in musl.
Even latest musl master de7dc1318f49 doesn't work for me with the same problem. The only way to make it work is to revert d1395c43c019 in musl, even with latest master.
link detection should be a kernel thing, or not? So how can a user-space change affect the kernel's link detection?
Any ideas?
Cheers,
Andre'

André Draszik

2019-01-21 16:52 UTC #4 lore raw

Thanks Khem,

So is it actually user-space that does link state detection? I thought it's the kernel.

I use connman here, I'll have a look, maybe some memory corruption causing this so...

Thanks again,
Andre'

Quoted 75 lines · Khem Raj

Hi André
this feature was implemented but does not emulate full glibc behavior and its possible that app is expecting the glibc behavior, I have fixed nfsutils app in past see for details
https://patchwork.kernel.org/patch/10707759/
I think you can start with suspecting app, I dont rule out libc bug but thats less likely
Quoted 63 lines · André Draszik
Hi,
Quoted 16 lines · Khem Raj
Complete changelogs are here
https://git.musl-libc.org/cgit/musl/log/?qt=range&q=c50985d5c8e316c5c464f352e79eeebfed1121a9..39ef612aa193cc6e954ac5a01574300ccd4b7ef9
Signed-off-by Khem Raj <[email protected]>
---
V2: move to latest master ( 3 more commits for regressions )
 meta/recipes-core/musl/musl_git.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/recipes- renamed from meta/recipes-core/musl/musl_git.bb +0 -0
core/musl/musl_git.bb
index 0d8f8eb2a4..b416ec45bf 100644
meta/recipes-core/musl/musl_git.bb line 7
Quoted 5 lines
meta/recipes-core/musl/musl_git.bb +1 -1
@@ -4,7 +4,7 @@
 require musl.inc
 inherit linuxloader
 
-SRCREV = "c50985d5c8e316c5c464f352e79eeebfed1121a9"
+SRCREV = "39ef612aa193cc6e954ac5a01574300ccd4b7ef9"
With this, link detection doesn't seem to work correctly anymore, instead
constant messages in the syslog:
Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver
[Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready
fec 30bf0000.ethernet eth1: Link is Up - 1Gbps/Full - flow control
rx/tx
IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver
[Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready
fec 30bf0000.ethernet eth1: Link is Up - 1Gbps/Full - flow control
rx/tx
IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready Atheros 8031 ethernet 30be0000.etherne:01: attached PHY driver
[Atheros 8031 ethernet] (mii_bus:phy_addr=30be0000.etherne:01, irq=-1)
etc.
I've tracked it down to commit d1395c43c019 ("allow freeaddrinfo of arbitrary sublists of addrinfo list" in musl.
Even latest musl master de7dc1318f49 doesn't work for me with the same problem. The only way to make it work is to revert d1395c43c019 in musl, even with latest master.
link detection should be a kernel thing, or not? So how can a user-space change affect the kernel's link detection?
Any ideas?
Cheers,
Andre'