On Mon, 13 May 2024 20:50:37 + Haiyang Zhang wrote:
> From the document, it can be:
> "ARM cores support both modes, but are most commonly used in, and typically
> default to little-endian mode. Most Linux distributions for ARM tend to be
> little-endian only."
>
On Mon, 13 May 2024 13:29:01 -0700 Haiyang Zhang wrote:
> - depends on PCI_MSI && X86_64
> + depends on PCI_MSI
> + depends on X86_64 || (ARM64 && !CPU_BIG_ENDIAN && ARM64_4K_PAGES)
Can ARM64 be big endian?
On Wed, 24 Apr 2024 03:33:37 -0700 Shradha Gupta wrote:
> Add sysfs attributes to read max_mtu and min_mtu value for
> network devices
Absolutely pointless. You posted v1, dumping this as a driver
specific value, even tho it's already reported by the core...
And you can't even produce a
On Tue, 2 Apr 2024 01:23:08 + Dexuan Cui wrote:
> > > I suggest the Fixes tag should be updated. Otherwise the fix
> > > looks good to me.
> >
> > Thanks for the suggestion. I actually thought about this before
> > submission.
> > I was worried about someone back ports the jumbo frame
On Thu, 14 Mar 2024 18:54:31 + Haiyang Zhang wrote:
> We understand irqbalance may be a "bad idea", and recommended some
> customers to disable it when having problems with it... But it's
> still enabled by default, and we cannot let all distro vendors
> and custom image makers to disable
On Wed, 13 Mar 2024 19:57:20 -0700 Shradha Gupta wrote:
> Default interrupts affinity for each queue:
>
> 25: 1103 02989138 Hyper-V PCIe MSI
> 4138200989697-edge mana_q0@pci:7870:00:00.0
> 26: 0 14005360 0 Hyper-V PCIe MSI
>
On Sun, 10 Mar 2024 21:19:50 -0700 Shradha Gupta wrote:
> > Seems unlikely, but if it does work we should enable it for all
> > devices, no driver by driver.
> You mean, if the usecase seems valid we should try to extend the framework
> mentioned by Rahul
>
On Fri, 8 Mar 2024 19:43:57 + Haiyang Zhang wrote:
> > Seems unlikely, but if it does work we should enable it for all
> > devices, no driver by driver.
> There are some existing drivers, like mlx, rmnet, netvsc, etc. using percpu
> counters. Are you suggesting we add a common API for all
On Fri, 8 Mar 2024 18:51:58 + Haiyang Zhang wrote:
> > Dynamic is a bit of an exaggeration, right? On a well-configured system
> > each CPU should use a single queue assigned thru XPS. And for manual
> > debug bpftrace should serve the purpose quite well.
>
> Some programs, like irqbalancer
On Fri, 8 Mar 2024 02:44:31 -0800 Shradha Gupta wrote:
> Extend 'ethtool -S' output for mana devices to include per-CPU packet
> stats
You haven't answered the questions on v1 and haven't addressed all
the feedback.
--
pw-bot: cr
pv-bot: feedback
On Thu, 7 Mar 2024 15:49:15 + Haiyang Zhang wrote:
> > > Extend 'ethtool -S' output for mana devices to include per-CPU packet
> > > stats
> >
> > But why? You already have per queue stats.
> Yes. But the q to cpu binding is dynamic, we also want the per-CPU stat
> to analyze the CPU
On Thu, 7 Mar 2024 06:52:12 -0800 Shradha Gupta wrote:
> Extend 'ethtool -S' output for mana devices to include per-CPU packet
> stats
But why? You already have per queue stats.
> Built-on: Ubuntu22
> Tested-on: Ubuntu22
I understand when people mention testing driver changes on particular
On Tue, 30 Jan 2024 22:09:57 -0800 Shradha Gupta wrote:
> This patch applies to net, which is missed in the subject. I will fix
> this in the new version of the patch. Thanks
$ git checkout net/main
[...]
$ git am raw
Applying: hv_netvsc:Register VF in netvsc_probe if NET_DEVICE_REGISTER missed
On Mon, 29 Jan 2024 23:18:55 -0800 Shradha Gupta wrote:
> If hv_netvsc driver is removed and reloaded, the NET_DEVICE_REGISTER
> handler cannot perform VF register successfully as the register call
> is received before netvsc_probe is finished. This is because we
> register
On Mon, 22 Jan 2024 08:00:59 -0800 Souradeep Chakrabarti wrote:
> IRQ node-numcore-num CPUperformance(%)
> 1 0 | 0 0 | 0 0 | 0-1 0
> 2 0 | 0 0 | 1 1 | 2-3 3
> 3 0 | 0 1 | 2 2 | 4-5 10
> 4 0 | 0 1 | 3 3 | 6-7
On Wed, 29 Nov 2023 14:17:39 -0800 Souradeep Chakrabarti wrote:
> On Mon, Nov 27, 2023 at 10:06:39AM -0800, Jakub Kicinski wrote:
> > On Mon, 27 Nov 2023 09:36:38 + Souradeep Chakrabarti wrote:
> > > easier to keep things inside the mana driver code here
> >
>
On Mon, 27 Nov 2023 09:36:38 + Souradeep Chakrabarti wrote:
> easier to keep things inside the mana driver code here
Easier for who? Upstream we care about consistency and maintainability
across all drivers.
On Tue, 21 Nov 2023 05:54:37 -0800 Souradeep Chakrabarti wrote:
> Existing MANA design assigns IRQ to every CPUs, including sibling
> hyper-threads
> in a core. This causes multiple IRQs to work on same CPU and may reduce the
> network
> performance with RSS.
>
> Improve the performance by
On Wed, 15 Nov 2023 08:14:06 -0800 Stephen Hemminger wrote:
> Jakub is right that in an ideal world, this could all be managed by
> userspace. But the management of network devices in Linux is a
> dumpster fire! Every distro invents there own solution, last time
> I counted there were six
On Fri, 10 Nov 2023 00:43:55 + Long Li wrote:
> The code above needs to work with and without netvsc (the possible
> master device) present.
I don't think that's a reasonable requirement for the kernel code.
The auto-bonding already puts the kernel into business of guessing
policy, which
On Tue, 7 Nov 2023 13:05:32 -0800 Haiyang Zhang wrote:
> If VF NIC is registered earlier, NETDEV_REGISTER event is replayed,
> but NETDEV_POST_INIT is not.
But Long Li sent the patch which starts to use POST_INIT against
the net-next tree. If we apply this to net and Long Li's patch to
net-next
On Wed, 8 Nov 2023 14:56:52 -0800 lon...@linuxonhyperv.com wrote:
> From: Long Li
>
> When a VF is being exposed form the kernel, it should be marked as "slave"
> before exposing to the user-mode. The VF is not usable without netvsc running
> as master. The user-mode should never see a VF
On Thu, 26 Oct 2023 14:22:34 -0700 Haiyang Zhang wrote:
> And, move register_netdevice_notifier() earlier, so the call back
> function is set before probing.
Are you sure you need this? I thought the netdev notifier "replays"
registration events (i.e. sends "fake" events for already present
On Fri, 27 Oct 2023 13:59:50 -0700 lon...@linuxonhyperv.com wrote:
> When a VF is being exposed form the kernel, it should be marked as "slave"
> before exposing to the user-mode. The VF is not usable without netvsc running
> as master. The user-mode should never see a VF without the "slave" flag.
On Fri, 27 Oct 2023 11:06:51 -0700 Haiyang Zhang wrote:
> From: Konstantin Taranov
>
> This patch uses a helper function for assignment of xdp_features.
> This change simplifies backports.
Generally making backports is not a strong enough reason to change
upstream code, but using the helper
On Thu, 26 Oct 2023 11:23:37 +0200 Przemek Kitszel wrote:
> this would now fit into one line
> (perhaps it's the same in other cases, I just checked this one manually)
I think cocci would fold lines automatically? Could be worth trying
spatch to do the conversion for that reason, if you aren't
26 matches
Mail list logo