DPDK patches and discussions
 help / color / mirror / Atom feed
From: Bruce Richardson <bruce.richardson@intel.com>
To: Fidaullah Noonari <fidaullah.noonari@emumba.com>
Cc: stephen@networkplumber.org, dev@dpdk.org,
	Fidaullah Noonari <fidaullah.noonari@gmai.com>
Subject: Re: [PATCH] usertools: add check for IOMMU support in dpdk-devbind
Date: Tue, 8 Mar 2022 12:56:55 +0000	[thread overview]
Message-ID: <YidSl5HGKMyF74Wo@bricha3-MOBL.ger.corp.intel.com> (raw)
In-Reply-To: <20220308124901.9838-1-fidaullah.noonari@gmai.com>

On Tue, Mar 08, 2022 at 05:49:01PM +0500, Fidaullah Noonari wrote:
> binding with vfio driver, when IOMMU is disabled, causes program to crash.
> this patch checks for IOMMU support, if it is disabled, changes vfio
> into unsafe noiommu mode and prints a warning message.
> 
> Signed-off-by: Fidaullah Noonari <fidaullah.noonari@gmai.com>
> ---
>  usertools/dpdk-devbind.py | 25 +++++++++++++++++++++++++
>  1 file changed, 25 insertions(+)
> 
> diff --git a/usertools/dpdk-devbind.py b/usertools/dpdk-devbind.py
> index ace4627218..f42eead0c4 100755
> --- a/usertools/dpdk-devbind.py
> +++ b/usertools/dpdk-devbind.py
> @@ -466,6 +466,26 @@ def unbind_all(dev_list, force=False):
>          unbind_one(d, force)
>  
>  
> +def check_iommu():
> +    """Check if IOMMU is enabled on system"""
> +    if len(os.listdir("/sys/class/iommu")) != 0:
> +        return True
> +    print("Warning: IOMMU support disabled")
> +    return False
> +
> +
> +def enable_noiommu_mode():
> +    """Enables the noiommu mode for vfio drivers"""
> +    filename = "/sys/module/vfio/parameters/enable_unsafe_noiommu_mode"
> +    try:
> +        f = open(filename, "w")
> +    except:
> +        sys.exit("Error: unable to enable noiommu mode, could not open '%s'" % filename)
> +    f.write("1")
> +    f.close()
> +    print("Warning: enabling unsafe no IOMMU mode for vfio drivers")
> +
> +
>  def bind_all(dev_list, driver, force=False):
>      """Bind method, takes a list of device locations"""
>      global devices
> @@ -492,6 +512,11 @@ def bind_all(dev_list, driver, force=False):
>      except ValueError as ex:
>          sys.exit(ex)
>  
> +    # check for IOMMU support
> +    if not check_iommu():
> +        if driver == "vfio-pci":
> +            enable_noiommu_mode()
> +
>      for d in dev_list:
>          bind_one(d, driver, force)
>  
I like the idea of doing this from the script, but I would rather it not
happen by default, since no-iommu reduces the security of the system. In
the normal case, I think it would be best if the script printed an error
message asking the user to enable iommu. We can also then add a flag to
allow the user to explicitly ask for no-iommu mode and take the risk.

/Bruce

  reply	other threads:[~2022-03-08 12:57 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-08 12:49 Fidaullah Noonari
2022-03-08 12:56 ` Bruce Richardson [this message]
2022-03-10  4:08   ` Fidaullah Noonari

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=YidSl5HGKMyF74Wo@bricha3-MOBL.ger.corp.intel.com \
    --to=bruce.richardson@intel.com \
    --cc=dev@dpdk.org \
    --cc=fidaullah.noonari@emumba.com \
    --cc=fidaullah.noonari@gmai.com \
    --cc=stephen@networkplumber.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).