<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body>
    Am 13.10.21 um 01:56 schrieb Sumit Semwal:<br>
    <blockquote type="cite" cite="mid:CAO_48GE_jCWY4waK-+FqVw5sbuoHddt4kWpnkpvyLDRC__yE+g@mail.gmail.com">
      
      <div dir="auto">
        <div>Hello Guangming, Christian,
          <div dir="auto"><br>
          </div>
          <br>
          <br>
          <div class="gmail_quote">
            <div dir="ltr" class="gmail_attr">On Tue, 12 Oct 2021, 14:09
              , <<a href="mailto:guangming.cao@mediatek.com" rel="noreferrer noreferrer noreferrer" target="_blank" moz-do-not-send="true">guangming.cao@mediatek.com</a>>
              wrote:<br>
            </div>
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex">From:
              Guangming Cao <<a href="mailto:Guangming.Cao@mediatek.com" rel="noreferrer
                noreferrer noreferrer noreferrer" target="_blank" moz-do-not-send="true">Guangming.Cao@mediatek.com</a>><br>
              <br>
              > Am 09.10.21 um 07:55 schrieb <a href="mailto:guangming.cao@mediatek.com" rel="noreferrer
                noreferrer noreferrer noreferrer" target="_blank" moz-do-not-send="true">guangming.cao@mediatek.com</a>:<br>
              > From: Guangming Cao <<a href="mailto:Guangming.Cao@mediatek.com" rel="noreferrer
                noreferrer noreferrer noreferrer" target="_blank" moz-do-not-send="true">Guangming.Cao@mediatek.com</a>><br>
              > ><br>
              > > If dma-buf don't want userspace users to touch
              the dmabuf buffer,<br>
              > > it seems we should add this restriction into
              dma_buf_ops.mmap,<br>
              > > not in this IOCTL:DMA_BUF_SET_NAME.<br>
              > ><br>
              > > With this restriction, we can only know the
              kernel users of the dmabuf<br>
              > > by attachments.<br>
              > > However, for many userspace users, such as
              userpsace users of dma_heap,<br>
              > > they also need to mark the usage of dma-buf, and
              they don't care about<br>
              > > who attached to this dmabuf, and seems it's no
              meaning to be waiting for<br>
              > > IOCTL:DMA_BUF_SET_NAME rather than mmap.<br>
              > <br>
              > Sounds valid to me, but I have no idea why this
              restriction was added in <br>
              > the first place.<br>
              > <br>
              > Can you double check the git history and maybe
              identify when that was <br>
              > added? Mentioning this change in the commit message
              then might make <br>
              > things a bit easier to understand.<br>
              > <br>
              > Thanks,<br>
              > Christian.<br>
              It was add in this patch: <a href="https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpatchwork.freedesktop.org%2Fpatch%2F310349%2F&data=04%7C01%7Cchristian.koenig%40amd.com%7C4149923e2b0646de82ce08d98ddbf2c2%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637696798278342557%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=N49RVF4s%2BGQ4D%2Ft1MOwRsCnslFnwobSB3G86pvP9m7A%3D&reserved=0" originalsrc="https://patchwork.freedesktop.org/patch/310349/" shash="rrFLG2SGWwCOnevYntPkTBSLrAKC6N1Tqo0D7JcYApk3UgzLEy90FuuUiWQy2kK2oLh/dD822UVt+oxD14gkAVIoBSSXLllKMcWgzyFWLFnQz/hYTHy36ByjwT0X8rljdha1Zk9DL+tA28zkfSnmyWzn6ZU0mcMBBfX5Ak2SJ/A=" rel="noreferrer noreferrer noreferrer noreferrer
                noreferrer" target="_blank" moz-do-not-send="true">https://patchwork.freedesktop.org/patch/310349/</a>.<br>
              However, there is no illustration about it.<br>
              I guess it wants users to set_name when no attachments on
              the dmabuf,<br>
              for case with attachments, we can find owner by device in
              attachments.<br>
              But just I said in commit message, this is might not a
              good idea.<br>
              <br>
              Do you have any idea?<br>
            </blockquote>
          </div>
        </div>
        <div dir="auto"><br>
        </div>
        <div dir="auto">For the original series, the idea was that
          allowing name change mid-use could confuse the users about the
          dma-buf. However, the rest of the series also makes sure each
          dma-buf have a unique inode, and any accounting should
          probably use that, without relying on the name as much.</div>
        <div dir="auto">So I don't have an objection to this change.</div>
      </div>
    </blockquote>
    <br>
    I suggest to add that explanation and the original commit id into
    the commit message.<br>
    <br>
    With that changed the patch has my rb as well.<br>
    <br>
    Regards,<br>
    Christian.<br>
    <br>
    <blockquote type="cite" cite="mid:CAO_48GE_jCWY4waK-+FqVw5sbuoHddt4kWpnkpvyLDRC__yE+g@mail.gmail.com">
      <div dir="auto">
        <div dir="auto"><br>
        </div>
        <div dir="auto">Best,</div>
        <div dir="auto">Sumit.</div>
        <div dir="auto">
          <div class="gmail_quote">
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex">
              > <br>
              > ><br>
              > > Signed-off-by: Guangming Cao <<a href="mailto:Guangming.Cao@mediatek.com" rel="noreferrer
                noreferrer noreferrer noreferrer" target="_blank" moz-do-not-send="true">Guangming.Cao@mediatek.com</a>><br>
              > > ---<br>
              > >   drivers/dma-buf/dma-buf.c | 14 ++------------<br>
              > >   1 file changed, 2 insertions(+), 12
              deletions(-)<br>
              > ><br>
              > > diff --git a/drivers/dma-buf/dma-buf.c
              b/drivers/dma-buf/dma-buf.c<br>
              > > index 511fe0d217a0..db2f4efdec32 100644<br>
              > > --- a/drivers/dma-buf/dma-buf.c<br>
              > > +++ b/drivers/dma-buf/dma-buf.c<br>
              > > @@ -325,10 +325,8 @@ static __poll_t
              dma_buf_poll(struct file *file, poll_table *poll)<br>
              > >   <br>
              > >   /**<br>
              > >    * dma_buf_set_name - Set a name to a specific
              dma_buf to track the usage.<br>
              > > - * The name of the dma-buf buffer can only be
              set when the dma-buf is not<br>
              > > - * attached to any devices. It could
              theoritically support changing the<br>
              > > - * name of the dma-buf if the same piece of
              memory is used for multiple<br>
              > > - * purpose between different devices.<br>
              > > + * It could theoretically support changing the
              name of the dma-buf if the same<br>
              > > + * piece of memory is used for multiple purpose
              between different devices.<br>
              > >    *<br>
              > >    * @dmabuf: [in]     dmabuf buffer that will
              be renamed.<br>
              > >    * @buf:    [in]     A piece of userspace
              memory that contains the name of<br>
              > > @@ -346,19 +344,11 @@ static long
              dma_buf_set_name(struct dma_buf *dmabuf, const char __user
              *buf)<br>
              > >     if (IS_ERR(name))<br>
              > >             return PTR_ERR(name);<br>
              > >   <br>
              > > -   dma_resv_lock(dmabuf->resv, NULL);<br>
              > > -   if
              (!list_empty(&dmabuf->attachments)) {<br>
              > > -           ret = -EBUSY;<br>
              > > -           kfree(name);<br>
              > > -           goto out_unlock;<br>
              > > -   }<br>
              > >     spin_lock(&dmabuf->name_lock);<br>
              > >     kfree(dmabuf->name);<br>
              > >     dmabuf->name = name;<br>
              > >     spin_unlock(&dmabuf->name_lock);<br>
              > >   <br>
              > > -out_unlock:<br>
              > > -   dma_resv_unlock(dmabuf->resv);<br>
              > >     return ret;<br>
              > >   }<br>
              > >   <br>
            </blockquote>
          </div>
        </div>
      </div>
    </blockquote>
    <br>
  </body>
</html>