Re: [PATCH] bfa: fix access to bfad_im_port_s

From: James Bottomley
Date: Wed Dec 06 2017 - 10:40:23 EST

Next message: Sumit Semwal: "Re: [PATCH 4.14 00/95] 4.14.4-stable review"
Previous message: Nicolas Pitre: "Re: [PATCH] kconfig: fix relational operators for bool and tristate symbols"
In reply to: Johannes Thumshirn: "Re: [PATCH] bfa: fix access to bfad_im_port_s"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Wed, 2017-12-06 at 09:07 +0100, Johannes Thumshirn wrote:
> James Bottomley <James.Bottomley@xxxxxxxxxxxxxxxxxxxxx> writes:
> >
> > On Tue, 2017-11-28 at 16:26 +0100, Johannes Thumshirn wrote:
> > >
> > > Commit 'cd21c605b2cf ("scsi: fc: provide fc_bsg_to_shost()
> > > helper")'
> > > changed access to bfa's 'struct bfad_im_port_s' by using
> > > shost_priv()
> > > instead of shost->hostdata[0].
> > >
> > > This lead to crashes like in the following back-trace:
> > >
> > > task: ffff880046375300 ti: ffff8800a2ef8000 task.ti:
> > > ffff8800a2ef8000
> > > RIP: e030:[<ffffffffa04c8252>]ÂÂ[<ffffffffa04c8252>]
> > > bfa_fcport_get_attr+0x82/0x260 [bfa]
> > > RSP: e02b:ffff8800a2efba10ÂÂEFLAGS: 00010046
> > > RAX: 575f415441536432 RBX: ffff8800a2efba28 RCX: 0000000000000000
> > > RDX: 0000000000000000 RSI: ffff8800a2efba28 RDI: ffff880004dc31d8
> > > RBP: ffff880004dc31d8 R08: 0000000000000000 R09: 0000000000000001
> > > R10: ffff88011fadc468 R11: 0000000000000001 R12: ffff880004dc31f0
> > > R13: 0000000000000200 R14: ffff880004dc61d0 R15: ffff880004947a10
> > > FS:ÂÂ00007feb1e489700(0000) GS:ffff88011fac0000(0000)
> > > knlGS:0000000000000000
> > > CS:ÂÂe033 DS: 0000 ES: 0000 CR0: 000000008005003b
> > > CR2: 00007ffe14e46c10 CR3: 00000000957b8000 CR4: 0000000000000660
> > > Stack:
> > > Âffff88001d4da000 ffff880004dc31c0 ffffffffa048a9df
> > > ffffffff81e56380
> > > Â0000000000000000 0000000000000000 0000000000000000
> > > 0000000000000000
> > > [] bfad_iocmd_ioc_get_info+0x4f/0x220 [bfa]
> > > [] bfad_iocmd_handler+0xa00/0xd40 [bfa]
> > > [] bfad_im_bsg_request+0xee/0x1b0 [bfa]
> > > [] fc_bsg_dispatch+0x10b/0x1b0 [scsi_transport_fc]
> > > [] bsg_request_fn+0x11d/0x1c0
> > > [] __blk_run_queue+0x2f/0x40
> > > [] blk_execute_rq_nowait+0xa8/0x160
> > > [] blk_execute_rq+0x77/0x120
> > > [] bsg_ioctl+0x1b6/0x200
> > > [] do_vfs_ioctl+0x2cd/0x4a0
> > > [] SyS_ioctl+0x74/0x80
> > > [] entry_SYSCALL_64_fastpath+0x12/0x6d
> > >
> > > Fixes: cd21c605b2cf ("scsi: fc: provide fc_bsg_to_shost()
> > > helper")
> > > Signed-off-by: Johannes Thumshirn <jthumshirn@xxxxxxx>
> > > Cc: Michal KoutnÃ <mkoutny@xxxxxxxx>
> > > ---
> > > Âdrivers/scsi/bfa/bfad_bsg.c | 6 ++++--
> > > Â1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/scsi/bfa/bfad_bsg.c
> > > b/drivers/scsi/bfa/bfad_bsg.c
> > > index 72ca2a2e08e2..09ef68c8225f 100644
> > > --- a/drivers/scsi/bfa/bfad_bsg.c
> > > +++ b/drivers/scsi/bfa/bfad_bsg.c
> > > @@ -3135,7 +3135,8 @@ bfad_im_bsg_vendor_request(struct bsg_job
> > > *job)
> > > Â struct fc_bsg_request *bsg_request = job->request;
> > > Â struct fc_bsg_reply *bsg_reply = job->reply;
> > > Â uint32_t vendor_cmd = bsg_request-
> > > >
> > > > rqst_data.h_vendor.vendor_cmd[0];
> > > - struct bfad_im_port_s *im_port =
> > > shost_priv(fc_bsg_to_shost(job));
> > > + struct Scsi_Host *shost = fc_bsg_to_shost(job);
> > > + struct bfad_im_port_s *im_port = shost->hostdata[0];
> > > Â struct bfad_s *bfad = im_port->bfad;
> > > Â void *payload_kbuf;
> > > Â int rc = -EINVAL;
> > > @@ -3350,7 +3351,8 @@ int
> > > Âbfad_im_bsg_els_ct_request(struct bsg_job *job)
> > > Â{
> > > Â struct bfa_bsg_data *bsg_data;
> > > - struct bfad_im_port_s *im_port =
> > > shost_priv(fc_bsg_to_shost(job));
> > > + struct Scsi_Host *shost = fc_bsg_to_shost(job);
> > > + struct bfad_im_port_s *im_port = shost->hostdata[0];
> > > Â struct bfad_s *bfad = im_port->bfad;
> > > Â bfa_bsg_fcpt_t *bsg_fcpt;
> > > Â struct bfad_fcxpÂÂÂÂ*drv_fcxp;
> >
> > OK, so we had a linux next failure with this:
> >
> > ÂÂÂÂAfter merging the scsi tree, today's linux-next build (x86_64
> > ÂÂÂÂallmodconfig) produced these warnings:
> >
> > ÂÂÂÂdrivers/scsi/bfa/bfad_bsg.c: In function
> > 'bfad_im_bsg_vendor_request':
> > ÂÂÂÂdrivers/scsi/bfa/bfad_bsg.c:3137:35: warning: initialization
> > makes pointer from integer without a cast [-Wint-conversion]
> > ÂÂÂÂÂÂstruct bfad_im_port_s *im_port = shost->hostdata[0];
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ^
> > ÂÂÂÂdrivers/scsi/bfa/bfad_bsg.c: In function
> > 'bfad_im_bsg_els_ct_request':
> > ÂÂÂÂdrivers/scsi/bfa/bfad_bsg.c:3353:35: warning: initialization
> > makes pointer from integer without a cast [-Wint-conversion]
> > ÂÂÂÂÂÂstruct bfad_im_port_s *im_port = shost->hostdata[0];
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ^
> >
> > ÂÂÂÂIntroduced by commit
> >
> > ÂÂÂÂÂÂ45349821ab3a ("scsi: bfa: fix access to bfad_im_port_s")
> >
> > ÂÂÂÂ--Â
> > ÂÂÂÂCheers,
> > ÂÂÂÂStephen Rothwell
> >
> > The reason is you should be using shost_priv(shost) not shost-
> > >hostdata[0].
>
>
> Yes, but using shost_priv() was resulting in the crash beforehand.
>
> shost_priv() doesn't do the same actually:
> static inline void *shost_priv(struct Scsi_Host *shost)
> {
> ÂÂÂÂÂÂÂÂreturn (void *)shost->hostdata;
> }
> vs
> ÂÂÂÂÂÂÂÂshost->hostdata[0];
> and:
> drivers/scsi/bfa/bfad_im.c:567: im_port->shost->hostdata[0] =
> (unsigned
> long)im_port;
>
> The quick fix would be:
>
> diff --git a/drivers/scsi/bfa/bfad_bsg.c
> b/drivers/scsi/bfa/bfad_bsg.c
> index 09ef68c8225f..075185db533f 100644
> --- a/drivers/scsi/bfa/bfad_bsg.c
> +++ b/drivers/scsi/bfa/bfad_bsg.c
> @@ -3136,7 +3136,7 @@ bfad_im_bsg_vendor_request(struct bsg_job *job)
> Â struct fc_bsg_reply *bsg_reply = job->reply;
> Â uint32_t vendor_cmd = bsg_request-
> >rqst_data.h_vendor.vendor_cmd[0];
> Â struct Scsi_Host *shost = fc_bsg_to_shost(job);
> - struct bfad_im_port_s *im_port = shost->hostdata[0];
> + struct bfad_im_port_s *im_port = (struct bfad_im_port_s *)
> shost->hostdata[0];
> Â struct bfad_s *bfad = im_port->bfad;
> Â void *payload_kbuf;
> Â int rc = -EINVAL;
> @@ -3352,7 +3352,7 @@ bfad_im_bsg_els_ct_request(struct bsg_job *job)
> Â{
> Â struct bfa_bsg_data *bsg_data;
> Â struct Scsi_Host *shost = fc_bsg_to_shost(job);
> - struct bfad_im_port_s *im_port = shost->hostdata[0];
> + struct bfad_im_port_s *im_port = (struct bfad_im_port_s *)
> shost->hostdata[0];
> Â struct bfad_s *bfad = im_port->bfad;
> Â bfa_bsg_fcpt_t *bsg_fcpt;
> Â struct bfad_fcxpÂÂÂÂ*drv_fcxp;
>
> Which is also consistent with the rest of the driver.

The crash before was because there was a level of indirection missing,
so what I meant something like

static inline struct bfad_import_s *
bfad_shost_priv(struct Scsi_Host *shost)
{
return *(struct bfad_im_port_s **)shost_priv(shost);
}

So we're using the correct primitives. ÂUsing a wrapper should ensure
the same mistake isn't made again.

James

Next message: Sumit Semwal: "Re: [PATCH 4.14 00/95] 4.14.4-stable review"
Previous message: Nicolas Pitre: "Re: [PATCH] kconfig: fix relational operators for bool and tristate symbols"
In reply to: Johannes Thumshirn: "Re: [PATCH] bfa: fix access to bfad_im_port_s"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]