From: Michal Orzel <michal.orzel@amd.com>
To: Oleg Nikitenko <oleshiiwood@gmail.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>,
Julien Grall <julien@xen.org>,
"xen-devel@lists.xenproject.org" <xen-devel@lists.xenproject.org>,
Bertrand Marquis <bertrand.marquis@arm.com>,
Carlo Nonato <carlo.nonato@minervasys.tech>
Subject: Re: xen cache colors in ARM
Date: Fri, 21 Apr 2023 14:52:01 +0200 [thread overview]
Message-ID: <43f5fdaa-47c7-6ec9-c477-dac62a5bceae@amd.com> (raw)
In-Reply-To: <CA+SAi2upd1P=KzbQS2BpD5zr3+OA=mrq7JiC7Zou9XSEJ_OYhA@mail.gmail.com>
Hi Oleg,
On 21/04/2023 14:49, Oleg Nikitenko wrote:
>
>
>
> Hello Michal,
>
> I was not able to enable earlyprintk in the xen for now.
> I decided to choose another way.
> This is a xen's command line that I found out completely.
>
> (XEN) $$$$ console=dtuart dtuart=serial0 dom0_mem=1600M dom0_max_vcpus=2 dom0_vcpus_pin bootscrub=0 vwfi=native sched=null timer_slop=0
Yes, adding a printk() in Xen was also a good idea.
>
> So you are absolutely right about a command line.
> Now I am going to find out why xen did not have the correct parameters from the device tree.
Maybe you will find this document helpful:
https://github.com/Xilinx/xen/blob/xlnx_rebase_4.16/docs/misc/arm/device-tree/booting.txt
~Michal
>
> Regards,
> Oleg
>
> пт, 21 апр. 2023 г. в 11:16, Michal Orzel <michal.orzel@amd.com <mailto:michal.orzel@amd.com>>:
>
>
> On 21/04/2023 10:04, Oleg Nikitenko wrote:
> >
> >
> >
> > Hello Michal,
> >
> > Yes, I use yocto.
> >
> > Yesterday all day long I tried to follow your suggestions.
> > I faced a problem.
> > Manually in the xen config build file I pasted the strings:
> In the .config file or in some Yocto file (listing additional Kconfig options) added to SRC_URI?
> You shouldn't really modify .config file but if you do, you should execute "make olddefconfig" afterwards.
>
> >
> > CONFIG_EARLY_PRINTK
> > CONFIG_EARLY_PRINTK_ZYNQMP
> > CONFIG_EARLY_UART_CHOICE_CADENCE
> I hope you added =y to them.
>
> Anyway, you have at least the following solutions:
> 1) Run bitbake xen -c menuconfig to properly set early printk
> 2) Find out how you enable other Kconfig options in your project (e.g. CONFIG_COLORING=y that is not enabled by default)
> 3) Append the following to "xen/arch/arm/configs/arm64_defconfig":
> CONFIG_EARLY_PRINTK_ZYNQMP=y
>
> ~Michal
>
> >
> > Host hangs in build time.
> > Maybe I did not set something in the config build file ?
> >
> > Regards,
> > Oleg
> >
> > чт, 20 апр. 2023 г. в 11:57, Oleg Nikitenko <oleshiiwood@gmail.com <mailto:oleshiiwood@gmail.com> <mailto:oleshiiwood@gmail.com <mailto:oleshiiwood@gmail.com>>>:
> >
> > Thanks Michal,
> >
> > You gave me an idea.
> > I am going to try it today.
> >
> > Regards,
> > O.
> >
> > чт, 20 апр. 2023 г. в 11:56, Oleg Nikitenko <oleshiiwood@gmail.com <mailto:oleshiiwood@gmail.com> <mailto:oleshiiwood@gmail.com <mailto:oleshiiwood@gmail.com>>>:
> >
> > Thanks Stefano.
> >
> > I am going to do it today.
> >
> > Regards,
> > O.
> >
> > ср, 19 апр. 2023 г. в 23:05, Stefano Stabellini <sstabellini@kernel.org <mailto:sstabellini@kernel.org> <mailto:sstabellini@kernel.org <mailto:sstabellini@kernel.org>>>:
> >
> > On Wed, 19 Apr 2023, Oleg Nikitenko wrote:
> > > Hi Michal,
> > >
> > > I corrected xen's command line.
> > > Now it is
> > > xen,xen-bootargs = "console=dtuart dtuart=serial0 dom0_mem=1600M dom0_max_vcpus=2 dom0_vcpus_pin bootscrub=0 vwfi=native sched=null
> > > timer_slop=0 way_size=65536 xen_colors=0-3 dom0_colors=4-7";
> >
> > 4 colors is way too many for xen, just do xen_colors=0-0. There is no
> > advantage in using more than 1 color for Xen.
> >
> > 4 colors is too few for dom0, if you are giving 1600M of memory to Dom0.
> > Each color is 256M. For 1600M you should give at least 7 colors. Try:
> >
> > xen_colors=0-0 dom0_colors=1-8
> >
> >
> >
> > > Unfortunately the result was the same.
> > >
> > > (XEN) - Dom0 mode: Relaxed
> > > (XEN) P2M: 40-bit IPA with 40-bit PA and 8-bit VMID
> > > (XEN) P2M: 3 levels with order-1 root, VTCR 0x0000000080023558
> > > (XEN) Scheduling granularity: cpu, 1 CPU per sched-resource
> > > (XEN) Coloring general information
> > > (XEN) Way size: 64kB
> > > (XEN) Max. number of colors available: 16
> > > (XEN) Xen color(s): [ 0 ]
> > > (XEN) alternatives: Patching with alt table 00000000002cc690 -> 00000000002ccc0c
> > > (XEN) Color array allocation failed for dom0
> > > (XEN)
> > > (XEN) ****************************************
> > > (XEN) Panic on CPU 0:
> > > (XEN) Error creating domain 0
> > > (XEN) ****************************************
> > > (XEN)
> > > (XEN) Reboot in five seconds...
> > >
> > > I am going to find out how command line arguments passed and parsed.
> > >
> > > Regards,
> > > Oleg
> > >
> > > ср, 19 апр. 2023 г. в 11:25, Oleg Nikitenko <oleshiiwood@gmail.com <mailto:oleshiiwood@gmail.com> <mailto:oleshiiwood@gmail.com <mailto:oleshiiwood@gmail.com>>>:
> > > Hi Michal,
> > >
> > > You put my nose into the problem. Thank you.
> > > I am going to use your point.
> > > Let's see what happens.
> > >
> > > Regards,
> > > Oleg
> > >
> > >
> > > ср, 19 апр. 2023 г. в 10:37, Michal Orzel <michal.orzel@amd.com <mailto:michal.orzel@amd.com> <mailto:michal.orzel@amd.com <mailto:michal.orzel@amd.com>>>:
> > > Hi Oleg,
> > >
> > > On 19/04/2023 09:03, Oleg Nikitenko wrote:
> > > >
> > > >
> > > >
> > > > Hello Stefano,
> > > >
> > > > Thanks for the clarification.
> > > > My company uses yocto for image generation.
> > > > What kind of information do you need to consult me in this case ?
> > > >
> > > > Maybe modules sizes/addresses which were mentioned by @Julien Grall <mailto:julien@xen.org <mailto:julien@xen.org> <mailto:julien@xen.org <mailto:julien@xen.org>>> ?
> > >
> > > Sorry for jumping into discussion, but FWICS the Xen command line you provided seems to be not the one
> > > Xen booted with. The error you are observing most likely is due to dom0 colors configuration not being
> > > specified (i.e. lack of dom0_colors=<> parameter). Although in the command line you provided, this parameter
> > > is set, I strongly doubt that this is the actual command line in use.
> > >
> > > You wrote:
> > > xen,xen-bootargs = "console=dtuart dtuart=serial0 dom0_mem=1600M dom0_max_vcpus=2 dom0_vcpus_pin bootscrub=0 vwfi=native
> > > sched=null timer_slop=0 way_szize=65536 xen_colors=0-3 dom0_colors=4-7";
> > >
> > > but:
> > > 1) way_szize has a typo
> > > 2) you specified 4 colors (0-3) for Xen, but the boot log says that Xen has only one:
> > > (XEN) Xen color(s): [ 0 ]
> > >
> > > This makes me believe that no colors configuration actually end up in command line that Xen booted with.
> > > Single color for Xen is a "default if not specified" and way size was probably calculated by asking HW.
> > >
> > > So I would suggest to first cross-check the command line in use.
> > >
> > > ~Michal
> > >
> > >
> > > >
> > > > Regards,
> > > > Oleg
> > > >
> > > > вт, 18 апр. 2023 г. в 20:44, Stefano Stabellini <sstabellini@kernel.org <mailto:sstabellini@kernel.org> <mailto:sstabellini@kernel.org <mailto:sstabellini@kernel.org>> <mailto:sstabellini@kernel.org <mailto:sstabellini@kernel.org> <mailto:sstabellini@kernel.org <mailto:sstabellini@kernel.org>>>>:
> > > >
> > > > On Tue, 18 Apr 2023, Oleg Nikitenko wrote:
> > > > > Hi Julien,
> > > > >
> > > > > >> This feature has not been merged in Xen upstream yet
> > > > >
> > > > > > would assume that upstream + the series on the ML [1] work
> > > > >
> > > > > Please clarify this point.
> > > > > Because the two thoughts are controversial.
> > > >
> > > > Hi Oleg,
> > > >
> > > > As Julien wrote, there is nothing controversial. As you are aware,
> > > > Xilinx maintains a separate Xen tree specific for Xilinx here:
> > > > https://github.com/xilinx/xen <https://github.com/xilinx/xen> <https://github.com/xilinx/xen <https://github.com/xilinx/xen>> <https://github.com/xilinx/xen <https://github.com/xilinx/xen> <https://github.com/xilinx/xen <https://github.com/xilinx/xen>>>
> > > >
> > > > and the branch you are using (xlnx_rebase_4.16) comes from there.
> > > >
> > > >
> > > > Instead, the upstream Xen tree lives here:
> > > > https://xenbits.xen.org/gitweb/?p=xen.git;a=summary <https://xenbits.xen.org/gitweb/?p=xen.git;a=summary> <https://xenbits.xen.org/gitweb/?p=xen.git;a=summary <https://xenbits.xen.org/gitweb/?p=xen.git;a=summary>> <https://xenbits.xen.org/gitweb/?p=xen.git;a=summary <https://xenbits.xen.org/gitweb/?p=xen.git;a=summary> <https://xenbits.xen.org/gitweb/?p=xen.git;a=summary <https://xenbits.xen.org/gitweb/?p=xen.git;a=summary>>>
> > > >
> > > >
> > > > The Cache Coloring feature that you are trying to configure is present
> > > > in xlnx_rebase_4.16, but not yet present upstream (there is an
> > > > outstanding patch series to add cache coloring to Xen upstream but it
> > > > hasn't been merged yet.)
> > > >
> > > >
> > > > Anyway, if you are using xlnx_rebase_4.16 it doesn't matter too much for
> > > > you as you already have Cache Coloring as a feature there.
> > > >
> > > >
> > > > I take you are using ImageBuilder to generate the boot configuration? If
> > > > so, please post the ImageBuilder config file that you are using.
> > > >
> > > > But from the boot message, it looks like the colors configuration for
> > > > Dom0 is incorrect.
> > > >
> > >
> > >
> > >
> >
>
next prev parent reply other threads:[~2023-04-21 12:52 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-18 8:43 xen cache colors in ARM Oleg Nikitenko
2023-04-18 8:59 ` Julien Grall
[not found] ` <CA+SAi2uPZ=Dq1GxF9Kj1zCO=nbb55ruVG31kH-TgdpR6bLznvA@mail.gmail.com>
[not found] ` <CA+SAi2s4WLiMEVa3u8rJRNZDpCpLTvnDygpObSUKxau-Q8dfyA@mail.gmail.com>
2023-04-18 11:09 ` Julien Grall
2023-04-18 11:26 ` Oleg Nikitenko
2023-04-18 11:29 ` Julien Grall
[not found] ` <CA+SAi2sZnrLzQoBn-e0GDy5De6PcGzxDCuJ3MSKicB_wB7o+Nw@mail.gmail.com>
2023-04-18 12:12 ` Fwd: " Oleg Nikitenko
2023-04-18 17:44 ` Stefano Stabellini
2023-04-19 7:03 ` Oleg Nikitenko
2023-04-19 7:37 ` Michal Orzel
2023-04-19 8:25 ` Oleg Nikitenko
2023-04-19 9:36 ` Oleg Nikitenko
2023-04-19 10:42 ` Michal Orzel
2023-04-19 20:05 ` Stefano Stabellini
2023-04-20 8:56 ` Oleg Nikitenko
2023-04-20 8:57 ` Oleg Nikitenko
2023-04-21 8:04 ` Oleg Nikitenko
2023-04-21 8:16 ` Michal Orzel
2023-04-21 12:49 ` Oleg Nikitenko
2023-04-21 12:52 ` Michal Orzel [this message]
2023-04-24 12:03 ` Oleg Nikitenko
2023-04-24 20:39 ` Stefano Stabellini
2023-04-25 7:33 ` Oleg Nikitenko
2023-04-25 18:20 ` Stefano Stabellini
2023-04-27 6:46 ` Oleg Nikitenko
2023-04-27 21:51 ` Stefano Stabellini
2023-05-05 8:28 ` Oleg Nikitenko
2023-05-05 8:34 ` Michal Orzel
2023-05-05 8:48 ` Oleg Nikitenko
2023-05-09 6:58 ` Oleg Nikitenko
2023-05-09 19:49 ` Stefano Stabellini
2023-05-11 10:02 ` Oleg Nikitenko
2023-05-11 10:15 ` Michal Orzel
2023-05-11 10:32 ` Oleg Nikitenko
2023-05-15 8:51 ` Oleg Nikitenko
2023-05-15 8:57 ` Michal Orzel
2023-05-16 12:15 ` Oleg Nikitenko
2023-05-16 14:40 ` Michal Orzel
2023-05-16 15:14 ` Oleg Nikitenko
2023-05-16 18:00 ` Michal Orzel
2023-05-19 13:38 ` Oleg Nikitenko
2023-05-19 13:51 ` Michal Orzel
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=43f5fdaa-47c7-6ec9-c477-dac62a5bceae@amd.com \
--to=michal.orzel@amd.com \
--cc=bertrand.marquis@arm.com \
--cc=carlo.nonato@minervasys.tech \
--cc=julien@xen.org \
--cc=oleshiiwood@gmail.com \
--cc=sstabellini@kernel.org \
--cc=xen-devel@lists.xenproject.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).