2024-06-28 13:59:17 |
bugproxy |
bug |
|
|
added bug |
2024-06-28 13:59:19 |
bugproxy |
tags |
|
architecture-s39064 bugnameltc-207082 severity-high targetmilestone-inin--- |
|
2024-06-28 13:59:19 |
bugproxy |
ubuntu: assignee |
|
Skipper Bug Screeners (skipper-screen-team) |
|
2024-06-28 13:59:44 |
bugproxy |
affects |
ubuntu |
linux (Ubuntu) |
|
2024-06-28 14:07:18 |
Frank Heimes |
bug task added |
|
ubuntu-z-systems |
|
2024-06-28 14:24:51 |
Frank Heimes |
ubuntu-z-systems: assignee |
|
Skipper Bug Screeners (skipper-screen-team) |
|
2024-06-28 14:24:53 |
Frank Heimes |
linux (Ubuntu): assignee |
Skipper Bug Screeners (skipper-screen-team) |
|
|
2024-06-28 14:25:01 |
Frank Heimes |
ubuntu-z-systems: importance |
Undecided |
High |
|
2024-07-03 08:50:50 |
Frank Heimes |
description |
Symptom:
Comparing Ubuntu 24.04 (kernelversion: 6.8.0-31-generic) against Ubuntu 22.04, all of our PCI-related network measurements on LPAR show massive throughput degradations (up to -72%). This shows for almost all workloads and numbers of connections, detereorating with the number of connections increasing. Especially drastic is the drop for a high number of parallel connections (50 and 250) and for small and medium-size transactional workloads. However, also for streaming-type workloads the degradation is clearly visible (up to 48% degradation).
Problem:
With kernel config setting CONFIG_IOMMU_DEFAULT_DMA_STRICT=y, IOMMU DMA mode changed from lazy to strict, causing these massive degradations.
Behavior can also be changed with a kernel commandline parameter (iommu.strict) for easy verification.
The issue is known and was quickly fixed upstream in December 2023, after being present for little less than two weeks.
Upstream fix: https://github.com/torvalds/linux/commit/b2b97a62f055dd638f7f02087331a8380d8f139a
Repro:
rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
0) Install uperf on both systems, client and server.
1) Start uperf at server: uperf -s
2) Start uperf at client: uperf -vai 5 -m uperf-profile.xml
3) Switch from strict to lazy mode using kernel commandline parameter iommu.strict=0.
4) Repeat steps 1) and 2).
Example:
For the following example, we chose the workload named above (rr1c-200x1000-250):
iommu.strict=1 (strict): 233464.914 TPS
iommu.strict=0 (lazy): 835123.193 TPS |
SRU Justification:
[Impact]
* With the introduction of c76c067e488c "s390/pci: Use dma-iommu layer"
(upstream with since kernel v6.7-rc1) there was a move (on s390x only)
to a different dma-iommu implementation.
* And with 92bce97f0c34 "s390/pci: Fix reset of IOMMU software counters"
(again upstream since 6.7(rc-1) the IOMMU_DEFAULT_DMA_LAZY kernel config
option should now be set to 'yes' by default for s390x.
* Since CONFIG_IOMMU_DEFAULT_DMA_STRICT and IOMMU_DEFAULT_DMA_LAZY
are related to each other CONFIG_IOMMU_DEFAULT_DMA_STRICT needs to be
set to "no" by default, which was upstream done by b2b97a62f055
"Revert "s390: update defconfigs"".
* These changes are all upstream, but were not picked up by the Ubuntu
kernel config.
* And not having these config options set properly is causing significant
PCI-related network throughput degradation (up to -72%).
* This shows for almost all workloads and numbers of connections,
deteriorating with the number of connections increasing.
* Especially drastic is the drop for a high number of parallel connections
(50 and 250) and for small and medium-size transactional workloads.
However, also for streaming-type workloads the degradation is clearly
visible (up to 48% degradation).
[Fix]
* The (upstream) fix is to set
IOMMU_DEFAULT_DMA_STRICT=no
and
IOMMU_DEFAULT_DMA_LAZY=y
(which is needed for the changed DAM IOMMU implementation since v6.7).
[Test Case]
* Setup two Ubuntu Server 24.04 LPARs (with kernel 6.8)
(one acting as server and as client)
that have (PCIe attached) RoCE Express devices attached
and that are connected to each other.
* Sample workload rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
* Install uperf on both systems, client and server.
* Start uperf at server: uperf -s
* Start uperf at client: uperf -vai 5 -m uperf-profile.xml
* Switch from strict to lazy mode
either using the new kernel (or the test build below)
or using kernel cmd-line parameter iommu.strict=0.
* Restart uperf on server and client, like before.
* Verification will be performed by IBM.
[Regression Potential]
* The is a certain regression potential, since the behavior with
the two modified kernel config options will change significantly.
* This may solve the (network) throughput issue with PCI devices,
but may also come with side-effects on other PCIe based devices
(the old compression adapters or the new NVMe carrier cards).
[Other]
* CCW devices are not affected.
* This is s390x-specific only, hence will not affect any other architecture.
__________
Symptom:
Comparing Ubuntu 24.04 (kernelversion: 6.8.0-31-generic) against Ubuntu 22.04, all of our PCI-related network measurements on LPAR show massive throughput degradations (up to -72%). This shows for almost all workloads and numbers of connections, detereorating with the number of connections increasing. Especially drastic is the drop for a high number of parallel connections (50 and 250) and for small and medium-size transactional workloads. However, also for streaming-type workloads the degradation is clearly visible (up to 48% degradation).
Problem:
With kernel config setting CONFIG_IOMMU_DEFAULT_DMA_STRICT=y, IOMMU DMA mode changed from lazy to strict, causing these massive degradations.
Behavior can also be changed with a kernel commandline parameter (iommu.strict) for easy verification.
The issue is known and was quickly fixed upstream in December 2023, after being present for little less than two weeks.
Upstream fix: https://github.com/torvalds/linux/commit/b2b97a62f055dd638f7f02087331a8380d8f139a
Repro:
rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
0) Install uperf on both systems, client and server.
1) Start uperf at server: uperf -s
2) Start uperf at client: uperf -vai 5 -m uperf-profile.xml
3) Switch from strict to lazy mode using kernel commandline parameter iommu.strict=0.
4) Repeat steps 1) and 2).
Example:
For the following example, we chose the workload named above (rr1c-200x1000-250):
iommu.strict=1 (strict): 233464.914 TPS
iommu.strict=0 (lazy): 835123.193 TPS |
|
2024-07-03 08:50:58 |
Frank Heimes |
ubuntu-z-systems: status |
New |
In Progress |
|
2024-07-03 08:51:03 |
Frank Heimes |
linux (Ubuntu): status |
New |
In Progress |
|
2024-07-03 08:56:16 |
Frank Heimes |
description |
SRU Justification:
[Impact]
* With the introduction of c76c067e488c "s390/pci: Use dma-iommu layer"
(upstream with since kernel v6.7-rc1) there was a move (on s390x only)
to a different dma-iommu implementation.
* And with 92bce97f0c34 "s390/pci: Fix reset of IOMMU software counters"
(again upstream since 6.7(rc-1) the IOMMU_DEFAULT_DMA_LAZY kernel config
option should now be set to 'yes' by default for s390x.
* Since CONFIG_IOMMU_DEFAULT_DMA_STRICT and IOMMU_DEFAULT_DMA_LAZY
are related to each other CONFIG_IOMMU_DEFAULT_DMA_STRICT needs to be
set to "no" by default, which was upstream done by b2b97a62f055
"Revert "s390: update defconfigs"".
* These changes are all upstream, but were not picked up by the Ubuntu
kernel config.
* And not having these config options set properly is causing significant
PCI-related network throughput degradation (up to -72%).
* This shows for almost all workloads and numbers of connections,
deteriorating with the number of connections increasing.
* Especially drastic is the drop for a high number of parallel connections
(50 and 250) and for small and medium-size transactional workloads.
However, also for streaming-type workloads the degradation is clearly
visible (up to 48% degradation).
[Fix]
* The (upstream) fix is to set
IOMMU_DEFAULT_DMA_STRICT=no
and
IOMMU_DEFAULT_DMA_LAZY=y
(which is needed for the changed DAM IOMMU implementation since v6.7).
[Test Case]
* Setup two Ubuntu Server 24.04 LPARs (with kernel 6.8)
(one acting as server and as client)
that have (PCIe attached) RoCE Express devices attached
and that are connected to each other.
* Sample workload rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
* Install uperf on both systems, client and server.
* Start uperf at server: uperf -s
* Start uperf at client: uperf -vai 5 -m uperf-profile.xml
* Switch from strict to lazy mode
either using the new kernel (or the test build below)
or using kernel cmd-line parameter iommu.strict=0.
* Restart uperf on server and client, like before.
* Verification will be performed by IBM.
[Regression Potential]
* The is a certain regression potential, since the behavior with
the two modified kernel config options will change significantly.
* This may solve the (network) throughput issue with PCI devices,
but may also come with side-effects on other PCIe based devices
(the old compression adapters or the new NVMe carrier cards).
[Other]
* CCW devices are not affected.
* This is s390x-specific only, hence will not affect any other architecture.
__________
Symptom:
Comparing Ubuntu 24.04 (kernelversion: 6.8.0-31-generic) against Ubuntu 22.04, all of our PCI-related network measurements on LPAR show massive throughput degradations (up to -72%). This shows for almost all workloads and numbers of connections, detereorating with the number of connections increasing. Especially drastic is the drop for a high number of parallel connections (50 and 250) and for small and medium-size transactional workloads. However, also for streaming-type workloads the degradation is clearly visible (up to 48% degradation).
Problem:
With kernel config setting CONFIG_IOMMU_DEFAULT_DMA_STRICT=y, IOMMU DMA mode changed from lazy to strict, causing these massive degradations.
Behavior can also be changed with a kernel commandline parameter (iommu.strict) for easy verification.
The issue is known and was quickly fixed upstream in December 2023, after being present for little less than two weeks.
Upstream fix: https://github.com/torvalds/linux/commit/b2b97a62f055dd638f7f02087331a8380d8f139a
Repro:
rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
0) Install uperf on both systems, client and server.
1) Start uperf at server: uperf -s
2) Start uperf at client: uperf -vai 5 -m uperf-profile.xml
3) Switch from strict to lazy mode using kernel commandline parameter iommu.strict=0.
4) Repeat steps 1) and 2).
Example:
For the following example, we chose the workload named above (rr1c-200x1000-250):
iommu.strict=1 (strict): 233464.914 TPS
iommu.strict=0 (lazy): 835123.193 TPS |
SRU Justification:
[Impact]
* With the introduction of c76c067e488c "s390/pci: Use dma-iommu layer"
(upstream with since kernel v6.7-rc1) there was a move (on s390x only)
to a different dma-iommu implementation.
* And with 92bce97f0c34 "s390/pci: Fix reset of IOMMU software counters"
(again upstream since 6.7(rc-1) the IOMMU_DEFAULT_DMA_LAZY kernel config
option should now be set to 'yes' by default for s390x.
* Since CONFIG_IOMMU_DEFAULT_DMA_STRICT and IOMMU_DEFAULT_DMA_LAZY
are related to each other CONFIG_IOMMU_DEFAULT_DMA_STRICT needs to be
set to "no" by default, which was upstream done by b2b97a62f055
"Revert "s390: update defconfigs"".
* These changes are all upstream, but were not picked up by the Ubuntu
kernel config.
* And not having these config options set properly is causing significant
PCI-related network throughput degradation (up to -72%).
* This shows for almost all workloads and numbers of connections,
deteriorating with the number of connections increasing.
* Especially drastic is the drop for a high number of parallel connections
(50 and 250) and for small and medium-size transactional workloads.
However, also for streaming-type workloads the degradation is clearly
visible (up to 48% degradation).
[Fix]
* The (upstream accepted) fix is to set
IOMMU_DEFAULT_DMA_STRICT=no
and
IOMMU_DEFAULT_DMA_LAZY=y
(which is needed for the changed DAM IOMMU implementation since v6.7).
[Test Case]
* Setup two Ubuntu Server 24.04 LPARs (with kernel 6.8)
(one acting as server and as client)
that have (PCIe attached) RoCE Express devices attached
and that are connected to each other.
* Sample workload rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
* Install uperf on both systems, client and server.
* Start uperf at server: uperf -s
* Start uperf at client: uperf -vai 5 -m uperf-profile.xml
* Switch from strict to lazy mode
either using the new kernel (or the test build below)
or using kernel cmd-line parameter iommu.strict=0.
* Restart uperf on server and client, like before.
* Verification will be performed by IBM.
[Regression Potential]
* The is a certain regression potential, since the behavior with
the two modified kernel config options will change significantly.
* This may solve the (network) throughput issue with PCI devices,
but may also come with side-effects on other PCIe based devices
(the old compression adapters or the new NVMe carrier cards).
[Other]
* CCW devices are not affected.
* This is s390x-specific only, hence will not affect any other architecture.
__________
Symptom:
Comparing Ubuntu 24.04 (kernelversion: 6.8.0-31-generic) against Ubuntu 22.04, all of our PCI-related network measurements on LPAR show massive throughput degradations (up to -72%). This shows for almost all workloads and numbers of connections, detereorating with the number of connections increasing. Especially drastic is the drop for a high number of parallel connections (50 and 250) and for small and medium-size transactional workloads. However, also for streaming-type workloads the degradation is clearly visible (up to 48% degradation).
Problem:
With kernel config setting CONFIG_IOMMU_DEFAULT_DMA_STRICT=y, IOMMU DMA mode changed from lazy to strict, causing these massive degradations.
Behavior can also be changed with a kernel commandline parameter (iommu.strict) for easy verification.
The issue is known and was quickly fixed upstream in December 2023, after being present for little less than two weeks.
Upstream fix: https://github.com/torvalds/linux/commit/b2b97a62f055dd638f7f02087331a8380d8f139a
Repro:
rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
0) Install uperf on both systems, client and server.
1) Start uperf at server: uperf -s
2) Start uperf at client: uperf -vai 5 -m uperf-profile.xml
3) Switch from strict to lazy mode using kernel commandline parameter iommu.strict=0.
4) Repeat steps 1) and 2).
Example:
For the following example, we chose the workload named above (rr1c-200x1000-250):
iommu.strict=1 (strict): 233464.914 TPS
iommu.strict=0 (lazy): 835123.193 TPS |
|
2024-07-03 09:22:00 |
Frank Heimes |
linux (Ubuntu): assignee |
|
Canonical Kernel Team (canonical-kernel-team) |
|
2024-07-03 11:16:34 |
Frank Heimes |
description |
SRU Justification:
[Impact]
* With the introduction of c76c067e488c "s390/pci: Use dma-iommu layer"
(upstream with since kernel v6.7-rc1) there was a move (on s390x only)
to a different dma-iommu implementation.
* And with 92bce97f0c34 "s390/pci: Fix reset of IOMMU software counters"
(again upstream since 6.7(rc-1) the IOMMU_DEFAULT_DMA_LAZY kernel config
option should now be set to 'yes' by default for s390x.
* Since CONFIG_IOMMU_DEFAULT_DMA_STRICT and IOMMU_DEFAULT_DMA_LAZY
are related to each other CONFIG_IOMMU_DEFAULT_DMA_STRICT needs to be
set to "no" by default, which was upstream done by b2b97a62f055
"Revert "s390: update defconfigs"".
* These changes are all upstream, but were not picked up by the Ubuntu
kernel config.
* And not having these config options set properly is causing significant
PCI-related network throughput degradation (up to -72%).
* This shows for almost all workloads and numbers of connections,
deteriorating with the number of connections increasing.
* Especially drastic is the drop for a high number of parallel connections
(50 and 250) and for small and medium-size transactional workloads.
However, also for streaming-type workloads the degradation is clearly
visible (up to 48% degradation).
[Fix]
* The (upstream accepted) fix is to set
IOMMU_DEFAULT_DMA_STRICT=no
and
IOMMU_DEFAULT_DMA_LAZY=y
(which is needed for the changed DAM IOMMU implementation since v6.7).
[Test Case]
* Setup two Ubuntu Server 24.04 LPARs (with kernel 6.8)
(one acting as server and as client)
that have (PCIe attached) RoCE Express devices attached
and that are connected to each other.
* Sample workload rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
* Install uperf on both systems, client and server.
* Start uperf at server: uperf -s
* Start uperf at client: uperf -vai 5 -m uperf-profile.xml
* Switch from strict to lazy mode
either using the new kernel (or the test build below)
or using kernel cmd-line parameter iommu.strict=0.
* Restart uperf on server and client, like before.
* Verification will be performed by IBM.
[Regression Potential]
* The is a certain regression potential, since the behavior with
the two modified kernel config options will change significantly.
* This may solve the (network) throughput issue with PCI devices,
but may also come with side-effects on other PCIe based devices
(the old compression adapters or the new NVMe carrier cards).
[Other]
* CCW devices are not affected.
* This is s390x-specific only, hence will not affect any other architecture.
__________
Symptom:
Comparing Ubuntu 24.04 (kernelversion: 6.8.0-31-generic) against Ubuntu 22.04, all of our PCI-related network measurements on LPAR show massive throughput degradations (up to -72%). This shows for almost all workloads and numbers of connections, detereorating with the number of connections increasing. Especially drastic is the drop for a high number of parallel connections (50 and 250) and for small and medium-size transactional workloads. However, also for streaming-type workloads the degradation is clearly visible (up to 48% degradation).
Problem:
With kernel config setting CONFIG_IOMMU_DEFAULT_DMA_STRICT=y, IOMMU DMA mode changed from lazy to strict, causing these massive degradations.
Behavior can also be changed with a kernel commandline parameter (iommu.strict) for easy verification.
The issue is known and was quickly fixed upstream in December 2023, after being present for little less than two weeks.
Upstream fix: https://github.com/torvalds/linux/commit/b2b97a62f055dd638f7f02087331a8380d8f139a
Repro:
rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
0) Install uperf on both systems, client and server.
1) Start uperf at server: uperf -s
2) Start uperf at client: uperf -vai 5 -m uperf-profile.xml
3) Switch from strict to lazy mode using kernel commandline parameter iommu.strict=0.
4) Repeat steps 1) and 2).
Example:
For the following example, we chose the workload named above (rr1c-200x1000-250):
iommu.strict=1 (strict): 233464.914 TPS
iommu.strict=0 (lazy): 835123.193 TPS |
SRU Justification:
[Impact]
* With the introduction of c76c067e488c "s390/pci: Use dma-iommu layer"
(upstream with since kernel v6.7-rc1) there was a move (on s390x only)
to a different dma-iommu implementation.
* And with 92bce97f0c34 "s390/pci: Fix reset of IOMMU software counters"
(again upstream since 6.7(rc-1) the IOMMU_DEFAULT_DMA_LAZY kernel config
option should now be set to 'yes' by default for s390x.
* Since CONFIG_IOMMU_DEFAULT_DMA_STRICT and IOMMU_DEFAULT_DMA_LAZY
are related to each other CONFIG_IOMMU_DEFAULT_DMA_STRICT needs to be
set to "no" by default, which was upstream done by b2b97a62f055
"Revert "s390: update defconfigs"".
* These changes are all upstream, but were not picked up by the Ubuntu
kernel config.
* And not having these config options set properly is causing significant
PCI-related network throughput degradation (up to -72%).
* This shows for almost all workloads and numbers of connections,
deteriorating with the number of connections increasing.
* Especially drastic is the drop for a high number of parallel connections
(50 and 250) and for small and medium-size transactional workloads.
However, also for streaming-type workloads the degradation is clearly
visible (up to 48% degradation).
[Fix]
* The (upstream accepted) fix is to set
IOMMU_DEFAULT_DMA_STRICT=no
and
IOMMU_DEFAULT_DMA_LAZY=y
(which is needed for the changed DAM IOMMU implementation since v6.7).
[Test Case]
* Setup two Ubuntu Server 24.04 systems (with kernel 6.8)
(one acting as server and as client)
that have (PCIe attached) RoCE Express devices attached
and that are connected to each other.
* Verify if the the iommu_group type of the used PCI device is DMA-FQ:
cat /sys/bus/pci/devices/<device>\:00\:00.0/iommu_group/type
DMA-FQ
* Sample workload rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
* Install uperf on both systems, client and server.
* Start uperf at server: uperf -s
* Start uperf at client: uperf -vai 5 -m uperf-profile.xml
* Switch from strict to lazy mode
either using the new kernel (or the test build below)
or using kernel cmd-line parameter iommu.strict=0.
* Restart uperf on server and client, like before.
* Verification will be performed by IBM.
[Regression Potential]
* The is a certain regression potential, since the behavior with
the two modified kernel config options will change significantly.
* This may solve the (network) throughput issue with PCI devices,
but may also come with side-effects on other PCIe based devices
(the old compression adapters or the new NVMe carrier cards).
[Other]
* CCW devices are not affected.
* This is s390x-specific only, hence will not affect any other architecture.
__________
Symptom:
Comparing Ubuntu 24.04 (kernelversion: 6.8.0-31-generic) against Ubuntu 22.04, all of our PCI-related network measurements on LPAR show massive throughput degradations (up to -72%). This shows for almost all workloads and numbers of connections, detereorating with the number of connections increasing. Especially drastic is the drop for a high number of parallel connections (50 and 250) and for small and medium-size transactional workloads. However, also for streaming-type workloads the degradation is clearly visible (up to 48% degradation).
Problem:
With kernel config setting CONFIG_IOMMU_DEFAULT_DMA_STRICT=y, IOMMU DMA mode changed from lazy to strict, causing these massive degradations.
Behavior can also be changed with a kernel commandline parameter (iommu.strict) for easy verification.
The issue is known and was quickly fixed upstream in December 2023, after being present for little less than two weeks.
Upstream fix: https://github.com/torvalds/linux/commit/b2b97a62f055dd638f7f02087331a8380d8f139a
Repro:
rr1c-200x1000-250 with rr1c-200x1000-250.xml:
<?xml version="1.0"?>
<profile name="TCP_RR">
<group nprocs="250">
<transaction iterations="1">
<flowop type="connect" options="remotehost=<remote IP> protocol=tcp tcp_nodelay" />
</transaction>
<transaction duration="300">
<flowop type="write" options="size=200"/>
<flowop type="read" options="size=1000"/>
</transaction>
<transaction iterations="1">
<flowop type="disconnect" />
</transaction>
</group>
</profile>
0) Install uperf on both systems, client and server.
1) Start uperf at server: uperf -s
2) Start uperf at client: uperf -vai 5 -m uperf-profile.xml
3) Switch from strict to lazy mode using kernel commandline parameter iommu.strict=0.
4) Repeat steps 1) and 2).
Example:
For the following example, we chose the workload named above (rr1c-200x1000-250):
iommu.strict=1 (strict): 233464.914 TPS
iommu.strict=0 (lazy): 835123.193 TPS |
|
2024-07-04 16:23:41 |
Stefan Bader |
nominated for series |
|
Ubuntu Noble |
|
2024-07-04 16:23:41 |
Stefan Bader |
bug task added |
|
linux (Ubuntu Noble) |
|
2024-07-04 16:23:41 |
Stefan Bader |
nominated for series |
|
Ubuntu Oracular |
|
2024-07-04 16:23:41 |
Stefan Bader |
bug task added |
|
linux (Ubuntu Oracular) |
|
2024-07-04 16:23:59 |
Stefan Bader |
linux (Ubuntu Noble): importance |
Undecided |
Medium |
|
2024-07-04 16:23:59 |
Stefan Bader |
linux (Ubuntu Noble): status |
New |
Fix Committed |
|
2024-07-04 16:24:15 |
Stefan Bader |
linux (Ubuntu Oracular): importance |
Undecided |
Medium |
|
2024-07-04 16:24:28 |
Stefan Bader |
linux (Ubuntu Oracular): status |
In Progress |
Fix Committed |
|
2024-07-04 16:24:33 |
Stefan Bader |
linux (Ubuntu Noble): status |
Fix Committed |
In Progress |
|
2024-07-04 18:57:40 |
Stefan Bader |
linux (Ubuntu Noble): status |
In Progress |
Fix Committed |
|
2024-07-04 19:29:16 |
Frank Heimes |
ubuntu-z-systems: status |
In Progress |
Fix Committed |
|
2024-07-11 19:45:43 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 severity-high targetmilestone-inin--- |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin--- verification-needed-noble-linux |
|
2024-07-19 07:29:30 |
bugproxy |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin--- verification-needed-noble-linux |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done verification-done-noble verification-needed-noble-linux |
|
2024-08-02 15:32:35 |
Timo Aaltonen |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done verification-done-noble verification-needed-noble-linux |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done verification-done-noble verification-done-noble-linux |
|
2024-08-06 20:56:20 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done verification-done-noble verification-done-noble-linux |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-noble-linux-ibm-gt-tdx |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
linux (Ubuntu Noble): status |
Fix Committed |
Fix Released |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
cve linked |
|
2024-25742 |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
cve linked |
|
2024-35984 |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
cve linked |
|
2024-35990 |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
cve linked |
|
2024-35992 |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
cve linked |
|
2024-35997 |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
cve linked |
|
2024-36008 |
|
2024-08-08 09:47:07 |
Launchpad Janitor |
cve linked |
|
2024-36016 |
|
2024-08-08 10:12:01 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-noble-linux-ibm-gt-tdx |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx |
|
2024-08-08 10:14:59 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-raspi |
|
2024-08-08 10:18:30 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-raspi |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-raspi |
|
2024-08-08 11:15:17 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-raspi |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-raspi-realtime-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-raspi verification-needed-noble-linux-raspi-realtime |
|
2024-08-08 12:45:07 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-raspi-realtime-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-raspi verification-needed-noble-linux-raspi-realtime |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-lowlatency-v2 kernel-spammed-noble-linux-raspi-realtime-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-lowlatency verification-needed-noble-linux-raspi verification-needed-noble-linux-raspi-realtime |
|
2024-08-09 03:27:31 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-lowlatency-v2 kernel-spammed-noble-linux-raspi-realtime-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-lowlatency verification-needed-noble-linux-raspi verification-needed-noble-linux-raspi-realtime |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-ibm-6.8-v2 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-lowlatency-v2 kernel-spammed-noble-linux-raspi-realtime-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-ibm-6.8 verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-lowlatency verification-needed-noble-linux-raspi verification-needed-noble-linux-raspi-realtime |
|
2024-08-12 11:36:32 |
Ubuntu Kernel Bot |
tags |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-ibm-6.8-v2 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-lowlatency-v2 kernel-spammed-noble-linux-raspi-realtime-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-ibm-6.8 verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-lowlatency verification-needed-noble-linux-raspi verification-needed-noble-linux-raspi-realtime |
architecture-s39064 bugnameltc-207082 kernel-spammed-jammy-linux-ibm-6.8-v2 kernel-spammed-jammy-linux-lowlatency-hwe-6.8-v2 kernel-spammed-jammy-linux-riscv-6.8-v2 kernel-spammed-noble-linux-azure-v2 kernel-spammed-noble-linux-ibm-gt-tdx-v2 kernel-spammed-noble-linux-lowlatency-v2 kernel-spammed-noble-linux-raspi-realtime-v2 kernel-spammed-noble-linux-raspi-v2 kernel-spammed-noble-linux-v2 severity-high targetmilestone-inin2404 verification-done-noble verification-done-noble-linux verification-needed-jammy-linux-ibm-6.8 verification-needed-jammy-linux-lowlatency-hwe-6.8 verification-needed-jammy-linux-riscv-6.8 verification-needed-noble-linux-azure verification-needed-noble-linux-ibm-gt-tdx verification-needed-noble-linux-lowlatency verification-needed-noble-linux-raspi verification-needed-noble-linux-raspi-realtime |
|