Re: [RFC v3] tcp: Replace TCP buffer structure by an iovec array

[David Gibson <david@gibson.dropbear.id.au>]

[-- Attachment #1: Type: text/plain, Size: 4259 bytes --]

On Thu, Mar 21, 2024 at 11:47:45PM +1100, David Gibson wrote:
> On Thu, Mar 21, 2024 at 01:21:03PM +0100, Laurent Vivier wrote:
> > On 3/21/24 11:55, David Gibson wrote:
> > > On Thu, Mar 21, 2024 at 09:26:58AM +0100, Laurent Vivier wrote:
> > > > On 3/21/24 02:26, David Gibson wrote:
> > > > > On Wed, Mar 20, 2024 at 05:31:46PM +0100, Laurent Vivier wrote:
> > > > > > To be able to provide pointers to TCP headers and IP headers without
> > > > > > worrying about alignment in the structure, split the structure into
> > > > > > several arrays and point to each part of the frame using an iovec array.
> > > > > > 
> > > > > > Using iovec also allows us to simply ignore the first entry when the
> > > > > > vnet length header is not needed.
> > > > > > 
> > > > > > Signed-off-by: Laurent Vivier <lvivier@redhat.com>
> > > > > 
> > > > > [snip]
> > > > > > +static_assert(MSS4 <= sizeof(tcp4_payload[0].data));
> > > > > 
> > > > > This generates a clang-tidy warning, because apparently the C11
> > > > > version of static_assert() requires a message - making it optional is
> > > > > a C2x extension.
> > > > > 
> > > > > Laurent, I know you're having trouble getting the full testsuite to
> > > > > run (but some ideas on that later today, I hope).  But could you
> > > > > please add a "make cppcheck" and "make clang-tidy" to your pre-post
> > > > > routine.
> > > > > 
> > > > 
> > > > I'll do.
> > > > Thank you to have checked that.
> > > > Are there any other commands to run before to send?
> > > > Something like scripts/chekpatch.pl we have in QEMU?
> > > > 
> > > > I have always the same problem with the test suite, even after a "make
> > > > realclean" in test (see attachment)
> > > > How to debug?
> > > 
> > > Huh.  Guest kernel can't find root.  That's not a fault I was
> > > expecting (and alas, won't be helped by the patch I have in the
> > > works).
> > > 
> > > What's really weird is I don't see any signs of loading the initrd
> > > before that failure, and with the mbuto images we're using we should
> > > never *leave* the initrd.  Which kind of makes sense - if it's not
> > > seeing the initrd, then there is indeed no root to mount.
> > > 
> > > Can you gran the full qemu command line it's trying to use?
> > > 
> > My file mbuto seems to be empty:
> 
> Welp, that would explain it.
> 
> > $ cd test/
> > $ make
> > ./mbuto/mbuto -p ./passt.mem.mbuto -c lz4 -f mbuto.mem.img
> > Applying profile from file ./passt.mem.mbuto
> > depmod: WARNING: could not open modules.builtin.modinfo at
> > /tmp/tmp.0Xr88SNogx/lib/modules/6.7.9-200.fc39.x86_64: No such file or
> > directory
> 
> So now we have two mysteries.  1) Why is depmod failing here, and 2)
> why isn't depmod failing causing mbuto to fail, rather than generating
> nonsense output.

Can you get the output of running mbuto with sh -x?  That might shed
some light.

> 
> > Size: bin  4.9M lib   15M kmod  218k total   20M compressed  4.1k
> > KERNEL=/boot/vmlinuz-6.7.9-200.fc39.x86_64
> > INITRD=/home/lvivier/Projects/passt/test/mbuto.mem.img
> > initramfs mounted at: /tmp/tmp.0Xr88SNogx
> > $ ls -l /home/lvivier/Projects/passt/test/mbuto.mem.img
> > -rw-r--r--. 1 lvivier lvivier 50 Mar 21 13:17 /home/lvivier/Projects/passt/test/mbuto.mem.img
> > $ gzip -d -  < /home/lvivier/Projects/passt/test/mbuto.mem.img|hexdump -C
> > 00000000  30 37 30 37 30 31 30 30  30 30 30 30 30 30 30 30  |0707010000000000|
> > 00000010  30 30 30 30 30 30 30 30  30 30 30 30 30 30 30 30  |0000000000000000|
> > 00000020  30 30 30 30 30 30 30 30  30 30 30 30 30 31 30 30  |0000000000000100|
> > 00000030  30 30 30 30 30 30 30 30  30 30 30 30 30 30 30 30  |0000000000000000|
> > *
> > 00000060  30 30 30 30 30 42 30 30  30 30 30 30 30 30 54 52  |00000B00000000TR|
> > 00000070  41 49 4c 45 52 21 21 21  00 00 00 00 00 00 00 00  |AILER!!!........|
> > 00000080  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
> > *
> > 00000200
> > 
> > Thanks,
> > Laurent
> > 
> 



-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]