: Renumber PCI error enums to start at zero
Subsystems (2)
PATCH 1/2 : Renumber PCI error enums to start at zero
Greg, Andrew,
This patch fixes an annoying numbering mistake.
Please apply this (and the next patch).
--linas
Subject: [PATCH 1/2]: Renumber PCI error enums to start at zero
Renumber the PCI error enums to start at zero for "normal/online". This allows un-initialized pci channel state (which defaults to zero) to be interpreted as "normal". Add very simple routine to check state, just in case this ever has to be fiddled with again.
----
include/linux/pci.h | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
include/linux/pci.h +8 -3
@@ -87,13 +87,13 @@ typedef unsigned int __bitwise pci_channenum pci_channel_state {/* I/O channel is in normal state */ - pci_channel_io_normal = (__force pci_channel_state_t )1 , + pci_channel_io_normal = (__force pci_channel_state_t )0 ,/* I/O to channel is blocked */ - pci_channel_io_frozen = (__force pci_channel_state_t )2 , + pci_channel_io_frozen = (__force pci_channel_state_t )1 ,/* PCI card is dead */ - pci_channel_io_perm_failure = (__force pci_channel_state_t )3 , + pci_channel_io_perm_failure = (__force pci_channel_state_t )2 , };typedef unsigned short __bitwise pci_bus_flags_t ;
@@ -181,6 +181,11 @@ struct pci_dev {#define to_pci_dev (n) container_of(n, struct pci_dev, dev)#define for_each_pci_dev (d) while ((d = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, d)) != NULL) +static inline int pci_channel_offline (struct pci_dev *pdev) +{ +return (pdev->error_state != pci_channel_io_normal); +} +static inline struct pci_cap_saved_state *pci_find_saved_cap (struct pci_dev *pci_dev,char cap) {
PATCH 2/2 : Use newly defined PCI channel offline routine
Use newly minted routine to access the PCI channel state.
----
drivers/net/e1000/e1000_main.c | 2 +- drivers/net/ixgb/ixgb_main.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
drivers/net/e1000/e1000_main.c +1 -1
@@ -3449,7 +3449,7 @@ e1000_update_stats(struct e1000_adapter*/ if (adapter->link_speed ==0 )return ; -if (pdev->error_state && pdev->error_state != pci_channel_io_normal) +if (pci_channel_offline (pdev))return ;spin_lock_irqsave (&adapter->stats_lock, flags);
drivers/net/ixgb/ixgb_main.c +1 -1
@@ -1564,7 +1564,7 @@ ixgb_update_stats(struct ixgb_adapter *astruct pci_dev *pdev = adapter->pdev;/* Prevent stats update while adapter is being reset */ -if (pdev->error_state && pdev->error_state != pci_channel_io_normal) +if (pci_channel_offline (pdev))return ; if((netdev->flags & IFF_PROMISC) || (netdev->flags & IFF_ALLMULTI) ||
Re: [PATCH 2/2]: Use newly defined PCI channel offline routine
Quoted 3 lines · Linas Vepstas
Use newly minted routine to access the PCI channel state.
Signed-off-by Linas Vepstas <[email protected]>
ACK, thanks Linas.
If it doesn't get picked up I can stack it on my queue for netdev later.
Auke
Quoted 32 lines
----
drivers/net/e1000/e1000_main.c | 2 +- drivers/net/ixgb/ixgb_main.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)Index:
linux-2.6.19-git7/drivers/net/e1000/e1000_main.c
===================================================================drivers/net/e1000/e1000_main.c +1
-1@@ -3449,7 +3449,7 @@ e1000_update_stats(struct e1000_adapter*/ if (adapter->link_speed ==0 )return ; -if (pdev->error_state && pdev->error_state != pci_channel_io_normal) +if (pci_channel_offline (pdev))return ;spin_lock_irqsave (&adapter->stats_lock, flags);Index:
linux-2.6.19-git7/drivers/net/ixgb/ixgb_main.c
===================================================================drivers/net/ixgb/ixgb_main.c +1
-1@@ -1564,7 +1564,7 @@ ixgb_update_stats(struct ixgb_adapter *astruct pci_dev *pdev = adapter->pdev;/* Prevent stats update while adapter is being reset */ -if (pdev->error_state && pdev->error_state != pci_channel_io_normal) +if (pci_channel_offline (pdev))return ; if((netdev->flags & IFF_PROMISC) || (netdev->flags & IFF_ALLMULTI) ||
Quoted 6 lines · Linas Vepstas
Greg, Andrew,
This patch fixes an annoying numbering mistake.
Please apply this (and the next patch).
Quoted 8 lines
--linas
Subject: [PATCH 1/2]: Renumber PCI error enums to start at zero
Renumber the PCI error enums to start at zero for "normal/online". This allows un-initialized pci channel state (which defaults to zero) to be interpreted as "normal". Add very simple routine to check state, just in case this ever has to be fiddled with again.
No, as you have a specific type for this state, never test it against "zero". That just defeats the whole issue of having a special type for this state.
So you should not have to change the values of the enumerated type for the code to work properly. In fact, I would argue that we should not change it for this very reason :)
So, care to respin these without the enum changes to provide the new function and use it?
thanks,
greg k-h
Quoted 4 lines · Greg KH
Quoted 3 lines · Linas Vepstas
Subject: [PATCH 1/2]: Renumber PCI error enums to start at zero
Quoted 8 lines
Quoted 4 lines
Renumber the PCI error enums to start at zero for "normal/online". This allows un-initialized pci channel state (which defaults to zero) to be interpreted as "normal". Add very simple routine to check state, just in case this ever has to be fiddled with again.
No, as you have a specific type for this state, never test it against "zero". That just defeats the whole issue of having a special type for this state.
Yes, well, I guess that was my initial thinking, which is why it got coded that way. But "in real life", the value in the struct isn't initialized (thus taking a value of zero). Its not initialized in deference to the traditional idea that "just saying bzero() should be enough".
However, that turned the test for error into a dorky double test:if(pdev->error_state && pdev->error_state != pci_channel_io_normal) which struck me as lame.
So, I'll ask: is it better to test for (state!=0 && state!=1) or, to initialize pdev->error_state = pci_channel_io_normal; in the driver probe code?
--linas
Quoted 14 lines · Linas Vepstas
Quoted 4 lines · Greg KH
Quoted 3 lines · Linas Vepstas
Subject: [PATCH 1/2]: Renumber PCI error enums to start at zero
Quoted 8 lines
Quoted 4 lines
Renumber the PCI error enums to start at zero for "normal/online". This allows un-initialized pci channel state (which defaults to zero) to be interpreted as "normal". Add very simple routine to check state, just in case this ever has to be fiddled with again.
No, as you have a specific type for this state, never test it against "zero". That just defeats the whole issue of having a special type for this state.
Quoted 5 lines
Yes, well, I guess that was my initial thinking, which is why it got coded that way. But "in real life", the value in the struct isn't initialized (thus taking a value of zero). Its not initialized in deference to the traditional idea that "just saying
bzero()should be enough".
Then properly initialize the thing.
Quoted 3 lines
However, that turned the test for error into a dorky double test:
if(pdev->error_state && pdev->error_state != pci_channel_io_normal)which struck me as lame.
Again, don't test an explicit enumerated type against "0", that's just foolish. Why have the explicit type if you are going to do that? Does sparse complain about it? It should if it doesn't...
Quoted 3 lines
So, I'll ask: is it better to test for (state!=0 && state!=1) or, to initialize
pdev->error_state=pci_channel_io_normal; in the driver probe code?
Initialize the state in the probe. Then check against the enumerated type, not an integer value. Sparse should complain if you try to do that.
Or, if you really want to test against integers, rip out those types, otherwise they are useless as you keep trying to go around them...
thanks,
greg k-h