| Message ID | 1542043689-214071-1-git-send-email-john.garry@huawei.com |
|---|---|
| State | New |
| Headers | show |
| Series | scsi: libsas: Remove pcidev reference | expand |
On 12/11/2018 17:49, John Garry wrote: > On 12/11/2018 17:32, Joe Perches wrote: >> On Tue, 2018-11-13 at 01:28 +0800, John Garry wrote: >>> Not all host drivers are PCI drivers - like hisi_sas, which supports a >>> platform driver - so remove reference to "pcidev". >>> >>> The debug level is also downgraded to KERN_ERR for the same message. >> [] >>> diff --git a/drivers/scsi/libsas/sas_discover.c >>> b/drivers/scsi/libsas/sas_discover.c >> [] >>> @@ -186,7 +186,7 @@ int sas_notify_lldd_dev_found(struct >>> domain_device *dev) >>> >>> res = i->dft->lldd_dev_found(dev); >>> if (res) { >>> - printk("sas: driver on pcidev %s cannot handle " >>> + pr_err("sas: driver on host %s cannot handle " >>> "device %llx, error:%d\n", >> >> As a printk without a KERN_<LEVEL> is printed at whatever >> CONFIG_MESSAGE_LOGLEVEL_DEFAULT is set to (default: 4 and >> rarely unchanged), this is effectively upgraded from a >> KERN_WARNING to KERN_ERR. > > ah, I thought that it was printed always. > > So maybe I'll just leave as-is. I forgot to mention that checkpatch complains about using printk() - that's why I changed it. Thanks, John > >> >> >> >> . >> > > > > . >
On 12/11/2018 18:30, Joe Perches wrote: > On Mon, 2018-11-12 at 17:55 +0000, John Garry wrote: >> On 12/11/2018 17:49, John Garry wrote: >>> On 12/11/2018 17:32, Joe Perches wrote: >>>> On Tue, 2018-11-13 at 01:28 +0800, John Garry wrote: >>>>> Not all host drivers are PCI drivers - like hisi_sas, which supports a >>>>> platform driver - so remove reference to "pcidev". >>>>> >>>>> The debug level is also downgraded to KERN_ERR for the same message. >>>> [] >>>>> diff --git a/drivers/scsi/libsas/sas_discover.c >>>>> b/drivers/scsi/libsas/sas_discover.c >>>> [] >>>>> @@ -186,7 +186,7 @@ int sas_notify_lldd_dev_found(struct >>>>> domain_device *dev) >>>>> >>>>> res = i->dft->lldd_dev_found(dev); >>>>> if (res) { >>>>> - printk("sas: driver on pcidev %s cannot handle " >>>>> + pr_err("sas: driver on host %s cannot handle " >>>>> "device %llx, error:%d\n", >>>> >>>> As a printk without a KERN_<LEVEL> is printed at whatever >>>> CONFIG_MESSAGE_LOGLEVEL_DEFAULT is set to (default: 4 and >>>> rarely unchanged), this is effectively upgraded from a >>>> KERN_WARNING to KERN_ERR. >>> >>> ah, I thought that it was printed always. >>> >>> So maybe I'll just leave as-is. >> >> I forgot to mention that checkpatch complains about using printk() - >> that's why I changed it. > > I believe that always assigning a KERN_<LEVEL> > is good thing so I am not against this change. > > Describing the change a bit more correctly than > 'debug level downgraded' would be useful. > Right > My preference is a patch like this which > always prefixes "sas: " to each log message > by adding a #define pr_fmt and converts the > bare printks to pr_notice, and converts the > slightly odd sas_printk macro and uses with > a hidden KERN_NOTICE to pr_notice. > > --- > drivers/scsi/libsas/sas_discover.c | 9 ++++----- > drivers/scsi/libsas/sas_expander.c | 33 ++++++++++++++++----------------- > drivers/scsi/libsas/sas_init.c | 10 +++++----- > drivers/scsi/libsas/sas_internal.h | 5 ++++- > drivers/scsi/libsas/sas_phy.c | 10 +++++----- > drivers/scsi/libsas/sas_port.c | 3 +-- > 6 files changed, 35 insertions(+), 35 deletions(-) > > diff --git a/drivers/scsi/libsas/sas_discover.c b/drivers/scsi/libsas/sas_discover.c > index dde433aa59c2..193d7a1ebeb3 100644 > --- a/drivers/scsi/libsas/sas_discover.c > +++ b/drivers/scsi/libsas/sas_discover.c > @@ -128,7 +128,8 @@ static int sas_get_port_device(struct asd_sas_port *port) > SAS_FANOUT_EXPANDER_DEVICE); > break; > default: [ ... ] This all seems reasonable > } else { > diff --git a/drivers/scsi/libsas/sas_internal.h b/drivers/scsi/libsas/sas_internal.h > index 50e12d662ffe..263cb41853c2 100644 > --- a/drivers/scsi/libsas/sas_internal.h > +++ b/drivers/scsi/libsas/sas_internal.h > @@ -32,7 +32,10 @@ > #include <scsi/libsas.h> > #include <scsi/sas_ata.h> > > -#define sas_printk(fmt, ...) printk(KERN_NOTICE "sas: " fmt, ## __VA_ARGS__) > +#ifdef pr_fmt > +#undef pr_fmt > +#endif > +#define pr_fmt(fmt) "sas: " fmt Some other subsystem may try to include this header, and gets its message prefix overwritten. Just a consequence for doing something bad, right? > > #define SAS_DPRINTK(fmt, ...) printk(KERN_DEBUG "sas: " fmt, ## __VA_ARGS__) > > diff --git a/drivers/scsi/libsas/sas_phy.c b/drivers/scsi/libsas/sas_phy.c > index bf3e1b979ca6..d628c0e6289d 100644 > --- a/drivers/scsi/libsas/sas_phy.c > +++ b/drivers/scsi/libsas/sas_phy.c > @@ -122,11 +122,11 @@ static void sas_phye_shutdown(struct work_struct *work) > phy->enabled = 0; > ret = i->dft->lldd_control_phy(phy, PHY_FUNC_DISABLE, NULL); > if (ret) > - sas_printk("lldd disable phy%02d returned %d\n", > - phy->id, ret); > - } else > - sas_printk("phy%02d is not enabled, cannot shutdown\n", > - phy->id); > + pr_notice("lldd disable phy%02d returned %d\n", > + phy->id, ret); > + } else { > + pr_notice("phy%02d is not enabled, cannot shutdown\n", phy->id); > + } > } > > /* ---------- Phy class registration ---------- */ > diff --git a/drivers/scsi/libsas/sas_port.c b/drivers/scsi/libsas/sas_port.c > index fad23dd39114..f2a25cef85b7 100644 > --- a/drivers/scsi/libsas/sas_port.c > +++ b/drivers/scsi/libsas/sas_port.c > @@ -147,8 +147,7 @@ static void sas_form_port(struct asd_sas_phy *phy) > } > > if (i >= sas_ha->num_phys) { > - printk(KERN_NOTICE "%s: couldn't find a free port, bug?\n", > - __func__); > + pr_notice("%s: couldn't find a free port, bug?\n", __func__); > spin_unlock_irqrestore(&sas_ha->phy_port_lock, flags); > return; > } > > > > . >
On Mon, 2018-11-12 at 18:48 +0000, John Garry wrote: > On 12/11/2018 18:30, Joe Perches wrote: [] > > diff --git a/drivers/scsi/libsas/sas_internal.h b/drivers/scsi/libsas/sas_internal.h [] > > @@ -32,7 +32,10 @@ > > #include <scsi/libsas.h> > > #include <scsi/sas_ata.h> > > > > -#define sas_printk(fmt, ...) printk(KERN_NOTICE "sas: " fmt, ## __VA_ARGS__) > > +#ifdef pr_fmt > > +#undef pr_fmt > > +#endif > > +#define pr_fmt(fmt) "sas: " fmt > > Some other subsystem may try to include this header, and gets its > message prefix overwritten. Just a consequence for doing something bad, > right? Right. And as this file is internal to drivers/scsi/libsas that seems very unlikely to occur. It also might useful to use the common debugging mechanisms and convert SAS_DPRINTK to pr_debug which would use the same #define pr_fmt.
On 12/11/2018 18:58, Joe Perches wrote: >>> +#define pr_fmt(fmt) "sas: " fmt >> > >> > Some other subsystem may try to include this header, and gets its >> > message prefix overwritten. Just a consequence for doing something bad, >> > right? > Right. > > And as this file is internal to drivers/scsi/libsas > that seems very unlikely to occur. > > It also might useful to use the common debugging > mechanisms and convert SAS_DPRINTK to pr_debug > which would use the same #define pr_fmt. > OK, I will try to put this all together as a marginally wider scope tidy-up. Cheers, John
On Mon, 2018-11-12 at 19:31 +0000, John Garry wrote: > On 12/11/2018 18:58, Joe Perches wrote: > > > > +#define pr_fmt(fmt) "sas: " fmt > > > > > > > > Some other subsystem may try to include this header, and gets its > > > > message prefix overwritten. Just a consequence for doing something bad, > > > > right? > > Right. > > > > And as this file is internal to drivers/scsi/libsas > > that seems very unlikely to occur. > > > > It also might useful to use the common debugging > > mechanisms and convert SAS_DPRINTK to pr_debug > > which would use the same #define pr_fmt. > > > > OK, I will try to put this all together as a marginally wider scope tidy-up. Thanks. Another thing that could be done is to change the #define pr_fmt(fmt) to KBUILD_MODNAME as that would prefix "libsas" instead of just "sas". I think that would be better but I didn't do that as it should be in a separate patch.
On 12/11/2018 19:52, Joe Perches wrote: > On Mon, 2018-11-12 at 19:31 +0000, John Garry wrote: >> On 12/11/2018 18:58, Joe Perches wrote: >>>>> +#define pr_fmt(fmt) "sas: " fmt >>>>> >>>>> Some other subsystem may try to include this header, and gets its >>>>> message prefix overwritten. Just a consequence for doing something bad, >>>>> right? >>> Right. >>> >>> And as this file is internal to drivers/scsi/libsas >>> that seems very unlikely to occur. >>> >>> It also might useful to use the common debugging >>> mechanisms and convert SAS_DPRINTK to pr_debug >>> which would use the same #define pr_fmt. >>> >> We have almost 100 references to SAS_DPRINTK. So when we replace with pr_debug, we will frequently have to realign the indentation. I'm beginning to think it's not worth the churn to remove SAS_DPRINTK. However, with regards to using pr_debug(), there are loops to jump through to ensure any output at all: - ensure DEBUG is defined or DYNAMIC DEBUG enabled - ensure current loglevel is high enough Having said that, there are actually lots of logs in libsas where debug level is too low (at pr_debug), and these could be upgraded. So I think I will just go through the code and revise these levels. >> OK, I will try to put this all together as a marginally wider scope tidy-up. > > Thanks. > > Another thing that could be done is to change > the #define pr_fmt(fmt) to KBUILD_MODNAME as > that would prefix "libsas" instead of just "sas". > I think "sas" makes more sense, as this is the technology, as opposed to "libsas", which is just the software framework name. > I think that would be better but I didn't do that > as it should be in a separate patch. > Cheers, John
On Tue, 2018-11-13 at 13:38 +0000, John Garry wrote: > On 12/11/2018 19:52, Joe Perches wrote: > > On Mon, 2018-11-12 at 19:31 +0000, John Garry wrote: > > > On 12/11/2018 18:58, Joe Perches wrote: > > > > > > +#define pr_fmt(fmt) "sas: " fmt [] > > > > It also might useful to use the common debugging > > > > mechanisms and convert SAS_DPRINTK to pr_debug > > > > which would use the same #define pr_fmt. {} > We have almost 100 references to SAS_DPRINTK. So when we replace with > pr_debug, we will frequently have to realign the indentation. I'm > beginning to think it's not worth the churn to remove SAS_DPRINTK. or just convert the #define > However, with regards to using pr_debug(), there are loops to jump > through to ensure any output at all: > - ensure DEBUG is defined or DYNAMIC DEBUG enabled True > - ensure current loglevel is high enough That's already the case. > Having said that, there are actually lots of logs in libsas where debug > level is too low (at pr_debug), and these could be upgraded. > > So I think I will just go through the code and revise these levels. Nice, thanks
diff --git a/drivers/scsi/libsas/sas_discover.c b/drivers/scsi/libsas/sas_discover.c index 6b12c67..3428ea9 100644 --- a/drivers/scsi/libsas/sas_discover.c +++ b/drivers/scsi/libsas/sas_discover.c @@ -186,7 +186,7 @@ int sas_notify_lldd_dev_found(struct domain_device *dev) res = i->dft->lldd_dev_found(dev); if (res) { - printk("sas: driver on pcidev %s cannot handle " + pr_err("sas: driver on host %s cannot handle " "device %llx, error:%d\n", dev_name(sas_ha->dev), SAS_ADDR(dev->sas_addr), res);
Not all host drivers are PCI drivers - like hisi_sas, which supports a platform driver - so remove reference to "pcidev". The debug level is also downgraded to KERN_ERR for the same message. Signed-off-by: John Garry <john.garry@huawei.com> -- 1.9.1