virtio — archive

[PATCH 1/1] split-ring: clarify the field len in the used ring

From: Cornelia Huck <[email protected]>
To: [email protected]
Date: Mon, 29 Nov 2021 12:23:00 -0000

On Mon, Nov 29 2021, Stefan Hajnoczi <[email protected]> wrote: > On Fri, Nov 26, 2021 at 06:12:33PM +0100, Halil Pasic wrote: >> The current descriptor is misleading: "the descriptor chain which was > > s/current descriptor/current description/? Indeed; easy mistake to make when you look at descriptor chains :) > >> used" generally includes both the descriptors that map the device read >> only, and descriptors that map the device write only portions of the >> buffer described by the descriptor chain. The argument that "used" means >> "written to" does not stand because one has to "use" the descriptor >> chain even when the whole buffer is device read only. >> >> One can argue, that the most straightforward way to interpret the phrase >> "total length of that descriptor chain" (without context) like the >> length of the list is usually defined: i.e. like the number of >> descriptors that constitute the chain. This is clearly not what we want >> here. Another intuitive way to interpret "total length of that >> descriptor chain" is size of the buffer mapped by the descriptor chain. >> This is not what we want either. In fact such wrongful interpretations >> have caused bugs in the wild. >> >> On the other hand, the text below the listing that gets modified here >> clearly describes the semantics of \field{len}. So let us replace >> the ambiguous explanation in the listing, with a hopefully non-ambiguous >> one. >> >> Signed-off-by: Halil Pasic <[email protected]> >> --- >> split-ring.tex | 5 ++++- >> 1 file changed, 4 insertions(+), 1 deletion(-) >> >> diff --git a/split-ring.tex b/split-ring.tex >> index bfef62d..68dca07 100644 >> --- a/split-ring.tex >> +++ b/split-ring.tex >> @@ -402,7 +402,10 @@ \subsection{The Virtqueue Used Ring}\label{sec:Basic Facilities of a Virtio Devi >> struct virtq_used_elem { >> /* Index of start of used descriptor chain. */ >> le32 id; >> - /* Total length of the descriptor chain which was used (written to) */ >> + /* >> + * The number of bytes written into the device writable portion of >> + * the buffer described by the descriptor chain. >> + */ > > Spaces vs tabs. It looks like 8 spaces should be used here. Nod. > > Otherwise: > Reviewed-by: Stefan Hajnoczi <[email protected]> I can push that with the fixup as an editorial update, if everyone agrees. (I don't think we need a vote, as the semantics of the field are already described below.)