From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga05.intel.com ([192.55.52.43]:20186 "EHLO mga05.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750830AbdFOWxn (ORCPT ); Thu, 15 Jun 2017 18:53:43 -0400 Subject: Re: RFC: pwritev2 regression test for invalid flags To: Adhemerval Zanella , GNU C Library , Christoph Hellwig Cc: Stephen Bates , "linux-block@vger.kernel.org" , Christoph Hellwig , Jens Axboe References: From: Jon Derrick Message-ID: <43631a74-8934-6940-3eaa-e562e49c5cb2@intel.com> Date: Thu, 1 Jun 2017 11:52:25 -0600 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8 Sender: linux-block-owner@vger.kernel.org List-Id: linux-block@vger.kernel.org Hi Zanella, On 06/15/2017 04:10 PM, Adhemerval Zanella wrote: > After the issue with LO_HI_LONG definition on x86_64-linux-gnu, I planed to add > this patch to check the above patch for correct check for invalid flags (which > would also have show this issue with LO_HI_LONG being used on p{read,write}v2). > > However it seems to trigger what I think it is a kernel bug on version that > provides p{read,write}v, where preadv2 does fails with EOPNOTSUPP but > pwritev2 does not. For instance, on x86_64-linux-gnu-x32 and i686-linux-gnu > (4.10.0-21-generic/x86_64): It looks like the issue is that you are going through the compat_writev path, which for some reason discards the flags parameter. Can you apply the patch below to your kernel? diff --git a/fs/read_write.c b/fs/read_write.c index c4f88af..f77eb22 100644 --- a/fs/read_write.c +++ b/fs/read_write.c @@ -1210,7 +1210,7 @@ static size_t compat_writev(struct file *file, if (!(file->f_mode & FMODE_CAN_WRITE)) goto out; - ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos, 0); + ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos, flags); out: if (ret > 0)