lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87a5gf4qsi.fsf@redhat.com>
Date: Tue, 10 Sep 2024 14:49:33 +0200
From: Vitaly Kuznetsov <vkuznets@...hat.com>
To: Yan Zhao <yan.y.zhao@...el.com>
Cc: pbonzini@...hat.com, seanjc@...gle.com, kvm@...r.kernel.org, Yan Zhao
 <yan.y.zhao@...el.com>, Kevin Tian <kevin.tian@...el.com>,
 kraxel@...hat.com, maarten.lankhorst@...ux.intel.com, mripard@...nel.org,
 tzimmermann@...e.de, airlied@...il.com, daniel@...ll.ch,
 virtualization@...ts.linux.dev, dri-devel@...ts.freedesktop.org,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] drm/bochs: use devm_ioremap_wc() to map framebuffer

Yan Zhao <yan.y.zhao@...el.com> writes:

> Opt for devm_ioremap_wc() over devm_ioremap() when mapping the framebuffer.
>
> Using devm_ioremap() results in the VA being mapped with PAT=UC-, which
> considerably slows down drm_fb_memcpy(). In contrast, devm_ioremap_wc()
> maps the VA with PAT set to WC, leading to better performance on platforms
> where access to UC memory is much slower than WC memory.
>
> Here's the performance data measured in a guest on the physical machine
> "Sapphire Rapids XCC".
> With host KVM honors guest PAT memory types, the effective memory type
> for this framebuffer range is
> - WC when devm_ioremap_wc() is used
> - UC- when devm_ioremap() is used.
>
> The data presented is an average from 10 execution runs.
>
> Cycles: Avg cycles of executed bochs_primary_plane_helper_atomic_update()
>         from VM boot to GDM show up
> Cnt:    Avg cnt of executed bochs_primary_plane_helper_atomic_update()
>         from VM boot to GDM show up
> T:      Avg time of each bochs_primary_plane_helper_atomic_update().
>
>  -------------------------------------------------
> |            | devm_ioremap() | devm_ioremap_wc() |
> |------------|----------------|-------------------|
> |  Cycles    |    211.545M    |   0.157M          |
> |------------|----------------|-------------------|
> |  Cnt       |     142        |   1917            |
> |------------|----------------|-------------------|
> |  T         |    0.1748s     |   0.0004s         |
>  -------------------------------------------------
>
> Note:
> Following the rebase to [3], the previously reported GDM failure on the
> VGA device [1] can no longer be reproduced, thanks to the memory management
> improvements made in [2]. Despite this, I have proceeded to submit this
> patch because of the noticeable performance improvements it provides.
>
> Reported-by: Vitaly Kuznetsov <vkuznets@...hat.com>

FWIW, this patch (alone) resolves the observed issue, thanks!

Tested-by: Vitaly Kuznetsov <vkuznets@...hat.com>

I, however, share Paolo's concern around existing VMs which KVM's change
is effectively breaking.

> Closes: https://lore.kernel.org/all/87jzfutmfc.fsf@redhat.com/#t
> Cc: Sean Christopherson <seanjc@...gle.com>
> Cc: Paolo Bonzini <pbonzini@...hat.com>
> Cc: Kevin Tian <kevin.tian@...el.com>
> Cc: Thomas Zimmermann <tzimmermann@...e.de>
> Signed-off-by: Yan Zhao <yan.y.zhao@...el.com>
> Link: https://lore.kernel.org/all/87jzfutmfc.fsf@redhat.com/#t [1]
> Link: https://patchwork.freedesktop.org/series/138086 [2]
> Link: https://gitlab.freedesktop.org/drm/misc/kernel/-/tree/drm-misc-next [3]
> ---
> v2:
> - Rebased to the latest drm-misc-next branch. [2]
> - Updated patch log to match the base code.
>
> v1: https://lore.kernel.org/all/20240909051529.26776-1-yan.y.zhao@intel.com
> ---
>  drivers/gpu/drm/tiny/bochs.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/tiny/bochs.c b/drivers/gpu/drm/tiny/bochs.c
> index 69c5f65e9853..9055b1dd66df 100644
> --- a/drivers/gpu/drm/tiny/bochs.c
> +++ b/drivers/gpu/drm/tiny/bochs.c
> @@ -268,7 +268,7 @@ static int bochs_hw_init(struct bochs_device *bochs)
>  	if (!devm_request_mem_region(&pdev->dev, addr, size, "bochs-drm"))
>  		DRM_WARN("Cannot request framebuffer, boot fb still active?\n");
>  
> -	bochs->fb_map = devm_ioremap(&pdev->dev, addr, size);
> +	bochs->fb_map = devm_ioremap_wc(&pdev->dev, addr, size);
>  	if (bochs->fb_map == NULL) {
>  		DRM_ERROR("Cannot map framebuffer\n");
>  		return -ENOMEM;

-- 
Vitaly


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ