Skip to content

K
Kernel

openeuler / Kernel
1 年多 前同步成功

通知 8
Star 0
Fork 0
K
Kernel
切换分支/标签
  1. 03 6月, 2021 2 次提交
    • M
      nvme-tcp: allow selecting the network interface for connections · 3ede8f72
      由 Martin Belanger 提交于 
      In our application, we need a way to force TCP connections to go out a
specific IP interface instead of letting Linux select the interface
based on the routing tables.

Add the 'host-iface' option to allow specifying the interface to use.
When the option host-iface is specified, the driver uses the specified
interface to set the option SO_BINDTODEVICE on the TCP socket before
connecting.

This new option is needed in addtion to the existing host-traddr for
the following reasons:

Specifying an IP interface by its associated IP address is less
intuitive than specifying the actual interface name and, in some cases,
simply doesn't work. That's because the association between interfaces
and IP addresses is not predictable. IP addresses can be changed or can
change by themselves over time (e.g. DHCP). Interface names are
predictable [1] and will persist over time. Consider the following
configuration.

1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state ...
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 100.0.0.100/24 scope global lo
       valid_lft forever preferred_lft forever
2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc ...
    link/ether 08:00:27:21:65:ec brd ff:ff:ff:ff:ff:ff
    inet 100.0.0.100/24 scope global enp0s3
       valid_lft forever preferred_lft forever
3: enp0s8: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc ...
    link/ether 08:00:27:4f:95:5c brd ff:ff:ff:ff:ff:ff
    inet 100.0.0.100/24 scope global enp0s8
       valid_lft forever preferred_lft forever

The above is a VM that I configured with the same IP address
(100.0.0.100) on all interfaces. Doing a reverse lookup to identify the
unique interface associated with 100.0.0.100 does not work here. And
this is why the option host_iface is required. I understand that the
above config does not represent a standard host system, but I'm using
this to prove a point: "We can never know how users will configure
their systems". By te way, The above configuration is perfectly fine
by Linux.

The current TCP implementation for host_traddr performs a
bind()-before-connect(). This is a common construct to set the source
IP address on a TCP socket before connecting. This has no effect on how
Linux selects the interface for the connection. That's because Linux
uses the Weak End System model as described in RFC1122 [2]. On the other
hand, setting the Source IP Address has benefits and should be supported
by linux-nvme. In fact, setting the Source IP Address is a mandatory
FedGov requirement (e.g. connection to a RADIUS/TACACS+ server).
Consider the following configuration.

$ ip addr list dev enp0s8
3: enp0s8: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc ...
    link/ether 08:00:27:4f:95:5c brd ff:ff:ff:ff:ff:ff
    inet 192.168.56.101/24 brd 192.168.56.255 scope global enp0s8
       valid_lft 426sec preferred_lft 426sec
    inet 192.168.56.102/24 scope global secondary enp0s8
       valid_lft forever preferred_lft forever
    inet 192.168.56.103/24 scope global secondary enp0s8
       valid_lft forever preferred_lft forever
    inet 192.168.56.104/24 scope global secondary enp0s8
       valid_lft forever preferred_lft forever

Here we can see that several addresses are associated with interface
enp0s8. By default, Linux always selects the default IP address,
192.168.56.101, as the source address when connecting over interface
enp0s8. Some users, however, want the ability to specify a different
source address (e.g., 192.168.56.102, 192.168.56.103, ...). The option
host_traddr can be used as-is to perform this function.

In conclusion, I believe that we need 2 options for TCP connections.
One that can be used to specify an interface (host-iface). And one that
can be used to set the source address (host-traddr). Users should be
allowed to use one or the other, or both, or none. Of course, the
documentation for host_traddr will need some clarification. It should
state that when used for TCP connection, this option only sets the
source address. And the documentation for host_iface should say that
this option is only available for TCP connections.

References:
[1] https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/
[2] https://tools.ietf.org/html/rfc1122

Tested both IPv4 and IPv6 connections.
Signed-off-by: NMartin Belanger <martin.belanger@dell.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
      3ede8f72
    • A
      nvme: extend and modify the APST configuration algorithm · ebd8a93a
      由 Alexey Bogoslavsky 提交于 
      The algorithm that was used until now for building the APST configuration
table has been found to produce entries with excessively long ITPT
(idle time prior to transition) for devices declaring relatively long
entry and exit latencies for non-operational power states. This leads
to unnecessary waste of power and, as a result, failure to pass
mandatory power consumption tests on Chromebook platforms.

The new algorithm is based on two predefined ITPT values and two
predefined latency tolerances. Based on these values, as well as on
exit and entry latencies reported by the device, the algorithm looks
for up to 2 suitable non-operational power states to use as primary
and secondary APST transition targets. The predefined values are
supplied to the nvme driver as module parameters:

 - apst_primary_timeout_ms (default: 100)
 - apst_secondary_timeout_ms (default: 2000)
 - apst_primary_latency_tol_us (default: 15000)
 - apst_secondary_latency_tol_us (default: 100000)

The algorithm echoes the approach used by Intel's and Microsoft's drivers
on Windows. The specific default parameter values are also based on those
drivers. Yet, this patch doesn't introduce the ability to dynamically
regenerate the APST table in the event of switching the power source from
AC to battery and back. Adding this functionality may be considered in the
future. In the meantime, the timeouts and tolerances reflect a compromise
between values used by Microsoft for AC and battery scenarios.

In most NVMe devices the new algorithm causes them to implement a more
aggressive power saving policy. While beneficial in most cases, this
sometimes comes at the price of a higher IO processing latency in certain
scenarios as well as at the price of a potential impact on the drive's
endurance (due to more frequent context saving when entering deep non-
operational states). So in order to provide a fallback for systems where
these regressions cannot be tolerated, the patch allows to revert to
the legacy behavior by setting either apst_primary_timeout_ms or
apst_primary_latency_tol_us parameter to 0. Eventually (and possibly after
fine tuning the default values of the module parameters) the legacy behavior
can be removed.

TESTING.

The new algorithm has been extensively tested. Initially, simulations were
used to compare APST tables generated by old and new algorithms for a wide
range of devices. After that, power consumption, performance and latencies
were measured under different workloads on devices from multiple vendors
(WD, Intel, Samsung, Hynix, Kioxia). Below is the description of the tests
and the findings.

General observations.
The effect the patch has on the APST table varies depending on the entry and
exit latencies advertised by the devices. For some devices, the effect is
negligible (e.g. Kioxia KBG40ZNS), for some significant, making the
transitions to PS3 and PS4 much quicker (e.g. WD SN530, Intel 760P), or making
the sleep deeper, PS4 rather than PS3 after a similar amount of time (e.g.
SK Hynix BC511). For some devices (e.g. Samsung PM991) the effect is mixed:
the initial transition happens after a longer idle time, but takes the device
to a lower power state.

Workflows.
In order to evaluate the patch's effect on the power consumption and latency,
7 workflows were used for each device. The workflows were designed to test
the scenarios where significant differences between the old and new behaviors
are most likely. Each workflow was tested twice: with the new and with the
old APST table generation implementation. Power consumption, performance and
latency were measured in the process. The following workflows were used:
1) Consecutive write at the maximum rate with IO depth of 2, with no pauses
2) Repeated pattern of 1000 consecutive writes of 4K packets followed by 50ms
   idle time
3) Repeated pattern of 1000 consecutive writes of 4K packets followed by 150ms
   idle time
4) Repeated pattern of 1000 consecutive writes of 4K packets followed by 500ms
   idle time
5) Repeated pattern of 1000 consecutive writes of 4K packets followed by 1.5s
   idle time
6) Repeated pattern of 1000 consecutive writes of 4K packets followed by 5s
   idle time
7) Repeated pattern of a single random read of a 4K packet followed by 150ms
   idle time

Power consumption
Actual power consumption measurements produced predictable results in
accordance with the APST mechanism's theory of operation.
Devices with long entry and exit latencies such as WD SN530 showed huge
improvement on scenarios 4,5 and 6 of up to 62%. Devices such as Kioxia
KBG40ZNS where the resulting APST table looks virtually identical with
both legacy and new algorithms, showed little or no change in the average power
consumption on all workflows. Devices with extra short latencies such as
Samsung PM991 showed moderate increase in power consumption of up to 18% in
worst case scenarios.
In addition, on Intel and Samsung devices a more complex impact was observed
on scenarios 3, 4 and 7. Our understanding is that due to longer stay in deep
non-operational states between the writes the devices start performing background
operations leading to an increase of power consumption. With the old APST tables
part of these operations are delayed until the scenario is over and a longer idle
period begins, but eventually this extra power is consumed anyway.

Performance.
In terms of performance measured on sustained write or read scenarios, the
effect of the patch is minimal as in this case the device doesn't enter low power
states.

Latency
As expected, in devices where the patch causes a more aggressive power saving
policy (e.g. WD SN530, Intel 760P), an increase in latency was observed in
certain scenarios. Workflow number 7, specifically designed to simulate the
worst case scenario as far as latency is concerned, indeed shows a sharp
increase in average latency (~2ms -> ~53ms on Intel 760P and 0.6 -> 10ms on
WD SN530). The latency increase on other workloads and other devices is much
milder or non-existent.
Signed-off-by: NAlexey Bogoslavsky <alexey.bogoslavsky@wdc.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
      ebd8a93a
  3. 12 5月, 2021 1 次提交
  5. 04 5月, 2021 4 次提交
  7. 22 4月, 2021 5 次提交
  9. 15 4月, 2021 14 次提交
  11. 09 4月, 2021 1 次提交
  13. 06 4月, 2021 3 次提交
  15. 03 4月, 2021 10 次提交