Re: patch for 2.1.90 net/core/iovec et al
Linus Torvalds (torvalds@transmeta.com)
Tue, 24 Mar 1998 10:40:10 -0800 (PST)
On Tue, 24 Mar 1998, David S. Miller wrote:
>
> Bill, memcpy_toiovec() must return the amount data not copied
> successfully when a fault occurs. The reason is that tcp_do_recvmsg()
> in tcp.c must be able to move the "read sequence" byte count back
> should memcpy_toiovec() fail mid-stream. Browse the call to
> memcpy_toiovec() in there for details.
>
> So please change this back in your patches.
No, I don't see any need to do this. If some user reads the socket with a
bad memory area resulting in a EFAULT, we don't care _what_ happens.
Remember: the kernel is there to act for the user, it is NOT there to
cover up for the mistakes the user does. Covering up only hides the real
bugs and you get a bad system over all..
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu