Re: file corruption

[Patrick Welche <prlw1%cam.ac.uk@localhost>]

On Wed, Apr 26, 2017 at 10:35:09AM +0100, Patrick Welche wrote:
> On Tue, Apr 25, 2017 at 04:04:25PM +0100, Patrick Welche wrote:
> > On Mon, Apr 24, 2017 at 05:47:50PM +0100, Patrick Welche wrote:
> > > On Mon, Apr 24, 2017 at 05:13:45PM +0100, Robert Swindells wrote:
> > > > 
> > > > Patrick Welche <prlw1%cam.ac.uk@localhost> wrote:
> > > > >The "file corruption" problem seems to be a problem with re(4):
> > > > 
> > > > [snip]
> > > > 
> > > > >Instead of using
> > > > >
> > > > >re0 at pci3 dev 0 function 0: RealTek 8168/8111 PCIe Gigabit Ethernet (rev. 0x07)
> > > > >re0: interrupting at msi3 vec 0
> > > > >rgephy0 at re0 phy 7: RTL8169S/8110S/8211 1000BASE-T media interface, rev. 5
> > > > >
> > > > >(with ip4csum etc enabled)
> > > > >
> > > > >I tried a USB wireless dongle, and the same cvs checkout over urtwm0
> > > > >appears to be fine. (No change of kernel, no reboot.)
> > > > 
> > > > Are you using IPv4 or IPv6 ?
> > > > 
> > > > I don't have any suggestions on a fix, just that re(4) only seems to
> > > > have hardware acceleration for IPv4 so you will get a different path
> > > > through the network stack for the two protocols.
> > > 
> > > IPv4
> > > 
> > > I'll try -current next, given that Jared added a RTKQ_IM_HW quirk for
> > > precisely this model...
> > 
> > Switching off {ip,tcp,udp}4csum with the previously described setup didn't
> > help and neither did nyftp's 201704222230Z with the quirk.
> 
> Tried iwm(4) with the nyftp 201704222230Z generic kernel and also see
> bits of cvs protocol:
> 
> src/crypto/external/bsd/openssl/lib/libcrypto/arch/x86_64/sha1-x86_64.S:
> 
>         addl    %edx,%ebx
>         rorxl   $27,%ecx,%r12d
>         rorxl   $2,%ecx,%edx
>         andl    %esi,Mod-time 27 Jan 2017 23:00:50 -0000
> MT +updated
> MT text U 
> MT fname src/crypto/external/bsd/openssl/lib/libcrypto/man/openssl_ui_compat.3
> 
> 
> re and iwm are way faster than urtwn... Userland is all from the nyftp
> 201704222230Z.

Netbooting 7.1, a cvs co through re(4) appears to be clean.
Unfortunately I can't run 7.1, as it's a new computer which only
has USB3 ports, so no keyboard on 7.1.