Skip to contentSkip to conversation

scsi: fix users of SCSI_FC_ATTRS to depend on NET

PATCH In review Randy Dunlap 3 participants 6 messages Sep 16, 2014

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 scsi: fix users of SCSI_FC_ATTRS to depend on NET

From: Randy Dunlap <[email protected]>

There are other kconfig symbols which select SCSI_FC_ATTRS, so they also need to depend on NET to fix kconfig warnings and build errors:

warning: (LIBFC && SCSI_IBMVFC && SCSI_QLA_FC && SCSI_LPFC && ZFCP && SCSI_BFA_FC && SCSI_CHELSIO_FCOE && FUSION_FC) selects SCSI_FC_ATTRS which has unmet direct dependencies (SCSI && NET) drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_event':
drivers/scsi/scsi_transport_fc.c:543:7: error: 'scsi_nl_sock' undeclared (first use in this function)
drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_vendor_event':
drivers/scsi/scsi_transport_fc.c:611:7: error: 'scsi_nl_sock' undeclared (first use in this function)

---
 drivers/message/fusion/Kconfig |    2 +-
 drivers/scsi/Kconfig           |    9 +++++----
 drivers/scsi/csiostor/Kconfig  |    2 +-
 drivers/scsi/qla2xxx/Kconfig   |    2 +-
 4 files changed, 8 insertions(+), 7 deletions(-)
drivers/message/fusion/Kconfig +1 -1
@@ -28,7 +28,7 @@ config FUSION_SPI
 
 config FUSION_FC
 	tristate "Fusion MPT ScsiHost drivers for FC"
-	depends on PCI && SCSI
+	depends on PCI && SCSI && NET
 	select SCSI_FC_ATTRS
 	---help---
 	  SCSI HOST support for a Fiber Channel host adapters.
drivers/scsi/Kconfig +5 -4
@@ -585,6 +585,7 @@ config HYPERV_STORAGE
 
 config LIBFC
 	tristate "LibFC module"
+	depends on NET
 	select SCSI_FC_ATTRS
 	select CRC32
 	---help---
@@ -815,7 +816,7 @@ config SCSI_IBMVSCSI
 
 config SCSI_IBMVFC
 	tristate "IBM Virtual FC support"
-	depends on PPC_PSERIES && SCSI
+	depends on PPC_PSERIES && SCSI && NET
 	select SCSI_FC_ATTRS
 	help
 	  This is the IBM POWER Virtual FC Client
@@ -1265,7 +1266,7 @@ source "drivers/scsi/qla4xxx/Kconfig"
 
 config SCSI_LPFC
 	tristate "Emulex LightPulse Fibre Channel Support"
-	depends on PCI && SCSI
+	depends on PCI && SCSI && NET
 	select SCSI_FC_ATTRS
 	select CRC_T10DIF
 	help
@@ -1675,7 +1676,7 @@ config SCSI_SUNESP
 
 config ZFCP
 	tristate "FCP host bus adapter driver for IBM eServer zSeries"
-	depends on S390 && QDIO && SCSI
+	depends on S390 && QDIO && SCSI && NET
 	select SCSI_FC_ATTRS
 	help
           If you want to access SCSI devices attached to your IBM eServer
@@ -1703,7 +1704,7 @@ config SCSI_PM8001
 
 config SCSI_BFA_FC
 	tristate "Brocade BFA Fibre Channel Support"
-	depends on PCI && SCSI
+	depends on PCI && SCSI && NET
 	select SCSI_FC_ATTRS
 	help
 	  This bfa driver supports all Brocade PCIe FC/FCOE host adapters.
drivers/scsi/csiostor/Kconfig +1 -1
@@ -1,6 +1,6 @@
 config SCSI_CHELSIO_FCOE
 	tristate "Chelsio Communications FCoE support"
-	depends on PCI && SCSI
+	depends on PCI && SCSI && NET
 	select SCSI_FC_ATTRS
 	select FW_LOADER
 	help
drivers/scsi/qla2xxx/Kconfig +1 -1
@@ -1,6 +1,6 @@
 config SCSI_QLA_FC
 	tristate "QLogic QLA2XXX Fibre Channel Support"
-	depends on PCI && SCSI
+	depends on PCI && SCSI && NET
 	select SCSI_FC_ATTRS
 	select FW_LOADER
 	---help---
#1 lore raw
Quoted 25 lines · Randy Dunlap

From: Randy Dunlap <[email protected]>

There are other kconfig symbols which select SCSI_FC_ATTRS, so they also need to depend on NET to fix kconfig warnings and build errors:

warning: (LIBFC && SCSI_IBMVFC && SCSI_QLA_FC && SCSI_LPFC && ZFCP && SCSI_BFA_FC && SCSI_CHELSIO_FCOE && FUSION_FC) selects SCSI_FC_ATTRS which has unmet direct dependencies (SCSI && NET) drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_event':
drivers/scsi/scsi_transport_fc.c:543:7: error: 'scsi_nl_sock' undeclared (first use in this function)
drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_vendor_event':
drivers/scsi/scsi_transport_fc.c:611:7: error: 'scsi_nl_sock' undeclared (first use in this function)

---
  drivers/message/fusion/Kconfig |    2 +-
  drivers/scsi/Kconfig           |    9 +++++----
  drivers/scsi/csiostor/Kconfig  |    2 +-
  drivers/scsi/qla2xxx/Kconfig   |    2 +-
  4 files changed, 8 insertions(+), 7 deletions(-)

drivers/scsi/Kconfig config SCSI_SUNESP · line 1682

Quoted 8 lines
drivers/scsi/Kconfig +1 -1
@@ -1675,7 +1676,7 @@ config SCSI_SUNESP
 
  config ZFCP
  	tristate "FCP host bus adapter driver for IBM eServer zSeries"
-	depends on S390 && QDIO && SCSI
+	depends on S390 && QDIO && SCSI && NET
  	select SCSI_FC_ATTRS
  	help
            If you want to access SCSI devices attached to your IBM eServer

I think zfcp does not have any (direct) dependency on NET.

It looks like SCSI_FC_ATTRS selects SCSI_NETLINK (declaring scsi_nl_sock) and only depends on SCSI but not on NET. SCSI_NETLINK itself only selects NET but does not model its direct depencency on NET?

Do I understand correctly, that the only solution is to make all INdirect (transitive) dependencies explicit with the LLDDs instead of making the direct exploiter SCSI_NETLINK(?) depend on NET?

Signature · 10 lines
Mit freundlichen Grüßen / Kind regards
Steffen Maier

Linux on System z Development

IBM Deutschland Research & Development GmbH
Vorsitzende des Aufsichtsrats: Martina Koederitz
Geschaeftsfuehrung: Dirk Wittkopp
Sitz der Gesellschaft: Boeblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294
maintainer
#2 lore raw

From: Steffen Maier <[email protected]>
Date: Tue, 16 Sep 2014 16:56:55 +0200

Quoted 6 lines

I think zfcp does not have any (direct) dependency on NET.

It looks like SCSI_FC_ATTRS selects SCSI_NETLINK (declaring scsi_nl_sock) and only depends on SCSI but not on NET. SCSI_NETLINK itself only selects NET but does not model its direct depencency on NET?

SCSI_NETLINK now has:

depends on NET

maintainer
#3 lore raw
In reply to Randy Dunlap

From: Randy Dunlap <[email protected]>
Date: Mon, 15 Sep 2014 23:35:21 -0700

Quoted 6 lines

From: Randy Dunlap <[email protected]>

There are other kconfig symbols which select SCSI_FC_ATTRS, so they also need to depend on NET to fix kconfig warnings and build errors:

Quoted 8 lines

warning: (LIBFC && SCSI_IBMVFC && SCSI_QLA_FC && SCSI_LPFC && ZFCP && SCSI_BFA_FC && SCSI_CHELSIO_FCOE && FUSION_FC) selects SCSI_FC_ATTRS which has unmet direct dependencies (SCSI && NET) drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_event':
drivers/scsi/scsi_transport_fc.c:543:7: error: 'scsi_nl_sock' undeclared (first use in this function)
drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_vendor_event':
drivers/scsi/scsi_transport_fc.c:611:7: error: 'scsi_nl_sock' undeclared (first use in this function)

Randy, this is starting to get convoluted.

It's pretty much pointless for SCSI_FC_ATTRS to depend on NET itself if we have to explicitly place a NET dependency on every single user of SCSI_FC_ATTRS.

We have expressed a proper dependency for SCSI_FC_ATTRS only to have it forcefully bypassed by every single user because we 'select' it instead of using 'depends'.

Can we just change these 'select' operations on SCSI_FC_ATTRS to just be 'depends' instead?

I really wouldn't mind if select had the effect of force enabling the dependencies of the select'd Kconfig symbol. Then it really does what we use it for (keeping the user from having to know obscure dependencies just to enable the feature they want) without the ugly side effect of dependency bypassing.

So a "select" would do a recursive "select" on all the dependencies needed to turn on the select'd object.

I doubt it would even need to recurse often at all, the ways we use this is almost always to turn on some top level major piece of infrastructure.

Anyways, that's a longer term thing and we need to fix this now, any opinions on just using 'depend SCSI_FX_ATTRS' to fix this?

#4 lore raw
Quoted 15 lines · David Miller

From: Randy Dunlap <[email protected]>
Date: Mon, 15 Sep 2014 23:35:21 -0700

Quoted 12 lines

From: Randy Dunlap <[email protected]>

There are other kconfig symbols which select SCSI_FC_ATTRS, so they also need to depend on NET to fix kconfig warnings and build errors:

warning: (LIBFC && SCSI_IBMVFC && SCSI_QLA_FC && SCSI_LPFC && ZFCP && SCSI_BFA_FC && SCSI_CHELSIO_FCOE && FUSION_FC) selects SCSI_FC_ATTRS which has unmet direct dependencies (SCSI && NET) drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_event':
drivers/scsi/scsi_transport_fc.c:543:7: error: 'scsi_nl_sock' undeclared (first use in this function)
drivers/scsi/scsi_transport_fc.c: In function 'fc_host_post_vendor_event':
drivers/scsi/scsi_transport_fc.c:611:7: error: 'scsi_nl_sock' undeclared (first use in this function)

Quoted 4 lines
Quoted 2 lines

Randy, this is starting to get convoluted.

Yes, I'm not happy about it either.

Quoted 20 lines

It's pretty much pointless for SCSI_FC_ATTRS to depend on NET itself if we have to explicitly place a NET dependency on every single user of SCSI_FC_ATTRS.

We have expressed a proper dependency for SCSI_FC_ATTRS only to have it forcefully bypassed by every single user because we 'select' it instead of using 'depends'.

Can we just change these 'select' operations on SCSI_FC_ATTRS to just be 'depends' instead?

I really wouldn't mind if select had the effect of force enabling the dependencies of the select'd Kconfig symbol. Then it really does what we use it for (keeping the user from having to know obscure dependencies just to enable the feature they want) without the ugly side effect of dependency bypassing.

So a "select" would do a recursive "select" on all the dependencies needed to turn on the select'd object.

Quoted 6 lines

I doubt it would even need to recurse often at all, the ways we use this is almost always to turn on some top level major piece of infrastructure.

Anyways, that's a longer term thing and we need to fix this now, any opinions on just using 'depend SCSI_FX_ATTRS' to fix this?

typo: depends on SCSI_FC_ATTRS

No, I certainly have no objection and prefer depends over select anyway.

Signature · 1 line
~Randy
maintainer
#5 lore raw

From: Randy Dunlap <[email protected]>
Date: Tue, 16 Sep 2014 13:24:46 -0700

Quoted 3 lines

typo: depends on SCSI_FC_ATTRS

No, I certainly have no objection and prefer depends over select anyway.

Ok, then please send me a patch which fixes things that way.

Thanks!