From mboxrd@z Thu Jan 1 00:00:00 1970 From: YueHaibing Subject: Re: [PATCH -next] IB/qib: Add missing err handle for qib_user_sdma_rb_insert Date: Fri, 4 Jan 2019 14:08:57 +0800 Message-ID: References: <20181221021938.13784-1-yuehaibing@huawei.com> <20190102171224.GA26765@ziepe.ca> <20190102184050.GI5424@mtr-leonro.mtl.com> <20190102190740.GH26211@ziepe.ca> <20190102192211.GJ5424@mtr-leonro.mtl.com> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20190102192211.GJ5424@mtr-leonro.mtl.com> Sender: linux-kernel-owner@vger.kernel.org To: Leon Romanovsky , Jason Gunthorpe Cc: dennis.dalessandro@intel.com, mike.marciniszyn@intel.com, dledford@redhat.com, linux-kernel@vger.kernel.org, linux-rdma@vger.kernel.org List-Id: linux-rdma@vger.kernel.org On 2019/1/3 3:22, Leon Romanovsky wrote:> On Wed, Jan 02, 2019 at 12:07:40PM -0700, Jason Gunthorpe wrote: >> On Wed, Jan 02, 2019 at 08:40:50PM +0200, Leon Romanovsky wrote: >>> On Wed, Jan 02, 2019 at 10:12:24AM -0700, Jason Gunthorpe wrote: >>>> On Fri, Dec 21, 2018 at 10:19:38AM +0800, YueHaibing wrote: >>>>> It should goto err handle if qib_user_sdma_rb_insert fails, >>>>> other than success return. >>>>> >>>>> Fixes: 67810e8c3c01 ("RDMA/qib: Remove all occurrences of BUG_ON()") >>>>> Signed-off-by: YueHaibing >>>>> drivers/infiniband/hw/qib/qib_user_sdma.c | 2 ++ >>>>> 1 file changed, 2 insertions(+) >>>>> >>>>> diff --git a/drivers/infiniband/hw/qib/qib_user_sdma.c b/drivers/infiniband/hw/qib/qib_user_sdma.c >>>>> index 31c523b..e87c0a7 100644 >>>>> +++ b/drivers/infiniband/hw/qib/qib_user_sdma.c >>>>> @@ -237,6 +237,8 @@ qib_user_sdma_queue_create(struct device *dev, int unit, int ctxt, int sctxt) >>>>> >>>>> ret = qib_user_sdma_rb_insert(&qib_user_sdma_rb_root, >>>>> sdma_rb_node); >>>>> + if (ret == 0) >>>>> + goto err_rb; >>>>> } >>>> >>>> This doesn't look right, what about undoing the kmalloc directly >>>> above? >>> >>> Back then, I came to conclusion that qib_user_sdma_rb_insert() never >>> returns 0. Otherwise, Dennis would see that BUG_ON() for a long time >>> ago. >> >> It fails if the index is in the RB tree, which would indicate a >> programming bug. > > I tried to say that this programming bug doesn't exist in the reality. In view of this, leave it just as this. > >> >> The way to replace the BUG_ON is something like: >> >> if (WARN_ON(ret == 0)) { >> kfree() >> goto err_rb; >> } >> >> Jason