[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAAo+4rXpSK-LzrHgU0g-c3n=q3xB3j+-GtmUZOLqQ7S44yS_2Q@mail.gmail.com>
Date: Fri, 16 Jan 2026 16:31:25 +0800
From: Chengfeng Ye <dg573847474@...il.com>
To: Jinpu Wang <jinpu.wang@...os.com>
Cc: "James E . J . Bottomley" <James.Bottomley@...senpartnership.com>,
"Martin K . Petersen" <martin.petersen@...cle.com>, Bart Van Assche <bvanassche@....org>,
linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org,
Chengfeng Ye <cyeaa@...nect.ust.hk>
Subject: Re: [PATCH v2] scsi: pm8001: Fix data race in sysfs SAS address read
> As James commented, sas address is uniq, not something changing all the time.
I think maybe the more noteworthy case is sas address could be read
when it is only partially initialized.
It might happen because the sysfs is mounted by scsi_add_host() at
line 1196 of pm8001_init.c
during probe, prior to the sas address being initialized by the
pm8001_init_sas_add() called
at line 1208. The small window could allow sysfs read return
non-initialized (should be
zero) or partially initialized sas address (due to the non-atomic read/write)
> Do you see any real issue?
No, I just read the code and spotted this potential issue, and so I reported it.
Reference: https://github.com/torvalds/linux/blob/master/drivers/scsi/pm8001/pm8001_init.c#L1196
Best regards,
Chengfeng
Jinpu Wang <jinpu.wang@...os.com> 于2026年1月16日周五 13:46写道:
>
> On Thu, Jan 15, 2026 at 6:55 PM Chengfeng Ye <dg573847474@...il.com> wrote:
> >
> > From: Chengfeng Ye <cyeaa@...nect.ust.hk>
> >
> > Fix a data race where pm8001_ctl_host_sas_address_show() reads
> > pm8001_ha->sas_addr without synchronization while it can be written
> > from interrupt context in pm8001_mpi_get_nvmd_resp().
> >
> > The write path is already protected by pm8001_ha->lock (held by
> > process_oq() when calling pm8001_mpi_get_nvmd_resp()),
> > but the sysfs read path accesses the 8-byte SAS address without
> > any synchronization, allowing torn reads.
> >
> > Thread interleaving scenario:
> >
> > Thread A (sysfs read) | Thread B (interrupt context)
> > -------------------------------------+------------------------------------
> > pm8001_ctl_host_sas_address_show() |
> > |- read sas_addr[0..3] |
> > | process_oq()
> > | |- spin_lock_irqsave(&lock)
> > | |- process_one_iomb()
> > | | |- pm8001_mpi_get_nvmd_resp()
> > | | |- memcpy(sas_addr, new, 8)
> > | | /* writes all 8 bytes */
> > | |- spin_unlock_irqrestore(&lock)
> > |- read sas_addr[4..7] |
> > /* gets mix of old and new */ |
> >
> > Fix by protecting the sysfs read with the same pm8001_ha->lock
> > using guard(spinlock_irqsave) for automatic lock cleanup.
> >
> > Signed-off-by: Chengfeng Ye <cyeaa@...nect.ust.hk>
> As James commented, sas address is uniq, not something changing all the time.
> Do you see any real issue?
> > ---
> > V1 -> V2: Use guard instead of lock/unlock pair
> >
> > drivers/scsi/pm8001/pm8001_ctl.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/drivers/scsi/pm8001/pm8001_ctl.c b/drivers/scsi/pm8001/pm8001_ctl.c
> > index cbfda8c04e95..200ee6bbd413 100644
> > --- a/drivers/scsi/pm8001/pm8001_ctl.c
> > +++ b/drivers/scsi/pm8001/pm8001_ctl.c
> > @@ -311,6 +311,8 @@ static ssize_t pm8001_ctl_host_sas_address_show(struct device *cdev,
> > struct Scsi_Host *shost = class_to_shost(cdev);
> > struct sas_ha_struct *sha = SHOST_TO_SAS_HA(shost);
> > struct pm8001_hba_info *pm8001_ha = sha->lldd_ha;
> > +
> > + guard(spinlock_irqsave)(&pm8001_ha->lock);
> > return sysfs_emit(buf, "0x%016llx\n",
> > be64_to_cpu(*(__be64 *)pm8001_ha->sas_addr));
> > }
> > --
> > 2.25.1
> >
Powered by blists - more mailing lists