| Message ID | 20230412185602.1592629-2-eajames@linux.ibm.com |
|---|---|
| State | New |
| Headers | show |
| Series | fsi: core: Lock scan mutex for master index removal | expand |
On Wed, 12 Apr 2023 at 18:56, Eddie James <eajames@linux.ibm.com> wrote: > > If a master scan occurs while the master is being unregistered, > the devicecs may end up with incorrect and possibly duplicate names, typo: devices > resulting in kernel warnings. Ensure the master index isn't changed > outside of the scan mutex. > > Signed-off-by: Eddie James <eajames@linux.ibm.com> > --- > drivers/fsi/fsi-core.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/fsi/fsi-core.c b/drivers/fsi/fsi-core.c > index fcbf0469ce3f..18d4d68482d7 100644 > --- a/drivers/fsi/fsi-core.c > +++ b/drivers/fsi/fsi-core.c > @@ -1354,12 +1354,12 @@ EXPORT_SYMBOL_GPL(fsi_master_register); > > void fsi_master_unregister(struct fsi_master *master) > { > + mutex_lock(&master->scan_lock); The ida functions are supposed to not require locking, but protecting against the test and changing of ->idx makes sense. Do you want to add a Fixes: line? > if (master->idx >= 0) { > ida_simple_remove(&master_ida, master->idx); the ida_simple functions are depreciated, at some point we should replace them with ida_alloc/ida_free. > master->idx = -1; > } > > - mutex_lock(&master->scan_lock); > fsi_master_unscan(master); > mutex_unlock(&master->scan_lock); > device_unregister(&master->dev); > -- > 2.31.1 >
On 5/31/23 02:47, Joel Stanley wrote: > On Wed, 12 Apr 2023 at 18:56, Eddie James <eajames@linux.ibm.com> wrote: >> If a master scan occurs while the master is being unregistered, >> the devicecs may end up with incorrect and possibly duplicate names, > typo: devices Thanks... > >> resulting in kernel warnings. Ensure the master index isn't changed >> outside of the scan mutex. >> >> Signed-off-by: Eddie James <eajames@linux.ibm.com> >> --- >> drivers/fsi/fsi-core.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/drivers/fsi/fsi-core.c b/drivers/fsi/fsi-core.c >> index fcbf0469ce3f..18d4d68482d7 100644 >> --- a/drivers/fsi/fsi-core.c >> +++ b/drivers/fsi/fsi-core.c >> @@ -1354,12 +1354,12 @@ EXPORT_SYMBOL_GPL(fsi_master_register); >> >> void fsi_master_unregister(struct fsi_master *master) >> { >> + mutex_lock(&master->scan_lock); > The ida functions are supposed to not require locking, but protecting > against the test and changing of ->idx makes sense. > > Do you want to add a Fixes: line? Sure. > >> if (master->idx >= 0) { >> ida_simple_remove(&master_ida, master->idx); > the ida_simple functions are depreciated, at some point we should > replace them with ida_alloc/ida_free. OK, I'll see if it makes sense to do that now. Thanks! Eddie > >> master->idx = -1; >> } >> >> - mutex_lock(&master->scan_lock); >> fsi_master_unscan(master); >> mutex_unlock(&master->scan_lock); >> device_unregister(&master->dev); >> -- >> 2.31.1 >>
diff --git a/drivers/fsi/fsi-core.c b/drivers/fsi/fsi-core.c index fcbf0469ce3f..18d4d68482d7 100644 --- a/drivers/fsi/fsi-core.c +++ b/drivers/fsi/fsi-core.c @@ -1354,12 +1354,12 @@ EXPORT_SYMBOL_GPL(fsi_master_register); void fsi_master_unregister(struct fsi_master *master) { + mutex_lock(&master->scan_lock); if (master->idx >= 0) { ida_simple_remove(&master_ida, master->idx); master->idx = -1; } - mutex_lock(&master->scan_lock); fsi_master_unscan(master); mutex_unlock(&master->scan_lock); device_unregister(&master->dev);
If a master scan occurs while the master is being unregistered, the devicecs may end up with incorrect and possibly duplicate names, resulting in kernel warnings. Ensure the master index isn't changed outside of the scan mutex. Signed-off-by: Eddie James <eajames@linux.ibm.com> --- drivers/fsi/fsi-core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)