lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:	Thu, 22 Mar 2012 11:28:19 +0800
From:	Gavin Shan <shangw@...ux.vnet.ibm.com>
To:	Benjamin Herrenschmidt <benh@...nel.crashing.org>
Cc:	sfr@...b.auug.org.au, grant.likely@...retlab.ca,
	linux-kernel@...r.kernel.org, linux-next@...r.kernel.org,
	paulus@...ba.org, linuxppc-dev@...ts.ozlabs.org
Subject: Re: [PATCH] EEH: remove eeh device from OF node

>> Originally, the PCI sensitive OF node is tracing the eeh device
>> through struct device_node::edev. However, it was regarded as
>> bad idea.
>> 
>> The patch removes struct device_node::edev. In addition, the
>> global list of eeh devices is introduced, and do retrival of
>> eeh device according to the given OF node through the global
>> list.
>
>So I'm not too happy with that. The main problem I see is that
>the code -constantly- calls of_node_to_eeh_dev.
>

Yes, it's somewhat expensive.

>IE. On any MMIO that happens to return all 1's, we'll call
>eeh_check_failure() for example, which does that. In fact for a hot path
>it's pretty horrid, it will:
>
> - Do an address cache lookup to get the pci_dev
> - Get the device-node frokm the pci_dev
> - Lookup your list to get the eeh_dev
>
>Shouldn't we instead change the address cache to contain eeh_dev
>instead ? And if you prefer keeping pci_dev, then it shouldn't be hard
>to stick a pointer to the eeh_dev in there via either struct archdata or
>maybe platform_data.
>
>The EEH code still, even after your rework, constantly pass "dn"'s as
>argument to functions just to convert it back to a struct eeh_dev rather
>than trying to pass the eeh_dev.
>

I'm working on explicit PE support and hopefully, I can send out
the code next week for review. The basic idea is to introuce
struct eeh_pe and attached eeh_dev to the corresponding PE. In turn,
eeh core will use PE as parameter instead of current eeh_dev ;-)

>So turning that into a list lookup will slow things down to a crawl.
>Also your patch never seems to remove anything from your list, which
>doesn't look right vs. hotplug.
>
>I suggest we fix that with a two phase approach:
>
> 1- ASAP so we can still get that into 3.4, move the eeh_dev pointer to
>struct pci_dn instead of struct device_node. This structure is
>accessible directly via dn->data and is ppc specific, that will be a
>better temporary solution and and adding stuff to the generic struct
>device_node.
>

I've sent patch again it and please take a look when you have time.

> 2- Then, what we need to do is generalize the use of eeh_dev rather
>than device_node as the main object being worked on in the eeh layer and
>thus as the argument to most functions.
>
>Cheers,
>Ben.
>

Thanks,
Gavin

>> Signed-off-by: Gavin Shan <shangw@...ux.vnet.ibm.com>
>> ---
>>  arch/powerpc/include/asm/eeh.h           |    7 +++++++
>>  arch/powerpc/platforms/pseries/eeh_dev.c |   29 ++++++++++++++++++++++++++++-
>>  include/linux/of.h                       |   10 ----------
>>  3 files changed, 35 insertions(+), 11 deletions(-)
>> 
>> diff --git a/arch/powerpc/include/asm/eeh.h b/arch/powerpc/include/asm/eeh.h
>> index d60f998..591e0a1 100644
>> --- a/arch/powerpc/include/asm/eeh.h
>> +++ b/arch/powerpc/include/asm/eeh.h
>> @@ -56,6 +56,7 @@ struct eeh_dev {
>>  	struct pci_controller *phb;	/* Associated PHB		*/
>>  	struct device_node *dn;		/* Associated device node	*/
>>  	struct pci_dev *pdev;		/* Associated PCI device	*/
>> +	struct list_head list;		/* Form the global link list	*/
>>  };
>>  
>>  static inline struct device_node *eeh_dev_to_of_node(struct eeh_dev *edev)
>> @@ -115,6 +116,7 @@ extern int eeh_subsystem_enabled;
>>   */
>>  #define EEH_MAX_ALLOWED_FREEZES 5
>>  
>> +struct eeh_dev *eeh_dev_from_of_node(struct device_node *dn);
>>  void * __devinit eeh_dev_init(struct device_node *dn, void *data);
>>  void __devinit eeh_dev_phb_init_dynamic(struct pci_controller *phb);
>>  void __init eeh_dev_phb_init(void);
>> @@ -132,6 +134,11 @@ void eeh_add_device_tree_early(struct device_node *);
>>  void eeh_add_device_tree_late(struct pci_bus *);
>>  void eeh_remove_bus_device(struct pci_dev *);
>>  
>> +static inline struct eeh_dev *of_node_to_eeh_dev(struct device_node *dn)
>> +{
>> +	return eeh_dev_from_of_node(dn);
>> +}
>> +
>>  /**
>>   * EEH_POSSIBLE_ERROR() -- test for possible MMIO failure.
>>   *
>> diff --git a/arch/powerpc/platforms/pseries/eeh_dev.c b/arch/powerpc/platforms/pseries/eeh_dev.c
>> index f3aed7d..925d3a3 100644
>> --- a/arch/powerpc/platforms/pseries/eeh_dev.c
>> +++ b/arch/powerpc/platforms/pseries/eeh_dev.c
>> @@ -34,6 +34,7 @@
>>  #include <linux/export.h>
>>  #include <linux/gfp.h>
>>  #include <linux/init.h>
>> +#include <linux/list.h>
>>  #include <linux/kernel.h>
>>  #include <linux/pci.h>
>>  #include <linux/string.h>
>> @@ -41,6 +42,30 @@
>>  #include <asm/pci-bridge.h>
>>  #include <asm/ppc-pci.h>
>>  
>> +/* eeh device list */
>> +static LIST_HEAD(eeh_dev_list);
>> +
>> +/**
>> + * eeh_dev_from_of_node - Retrieve EEH device according to OF node
>> + * @dn: OF node
>> + *
>> + * All existing eeh devices have been put into the global list.
>> + * In addition, the eeh device is tracing the corresponding
>> + * OF node. The function is used to retrieve the corresponding
>> + * eeh device according to the given OF node.
>> + */
>> +struct eeh_dev *eeh_dev_from_of_node(struct device_node *dn)
>> +{
>> +	struct eeh_dev *edev = NULL;
>> +
>> +	list_for_each_entry(edev, &eeh_dev_list, list) {
>> +		if (edev->dn && edev->dn == dn)
>> +			return edev;
>> +	}
>> +
>> +	return NULL;
>> +}
>> +
>>  /**
>>   * eeh_dev_init - Create EEH device according to OF node
>>   * @dn: device node
>> @@ -62,10 +87,12 @@ void * __devinit eeh_dev_init(struct device_node *dn, void *data)
>>  	}
>>  
>>  	/* Associate EEH device with OF node */
>> -	dn->edev  = edev;
>>  	edev->dn  = dn;
>>  	edev->phb = phb;
>>  
>> +	/* Add to global list */
>> +	list_add_tail(&edev->list, &eeh_dev_list);
>> +
>>  	return NULL;
>>  }
>>  
>> diff --git a/include/linux/of.h b/include/linux/of.h
>> index 3e710d8..a75a831 100644
>> --- a/include/linux/of.h
>> +++ b/include/linux/of.h
>> @@ -58,9 +58,6 @@ struct device_node {
>>  	struct	kref kref;
>>  	unsigned long _flags;
>>  	void	*data;
>> -#if defined(CONFIG_EEH)
>> -	struct eeh_dev *edev;
>> -#endif
>>  #if defined(CONFIG_SPARC)
>>  	char	*path_component_name;
>>  	unsigned int unique_id;
>> @@ -75,13 +72,6 @@ struct of_phandle_args {
>>  	uint32_t args[MAX_PHANDLE_ARGS];
>>  };
>>  
>> -#if defined(CONFIG_EEH)
>> -static inline struct eeh_dev *of_node_to_eeh_dev(struct device_node *dn)
>> -{
>> -	return dn->edev;
>> -}
>> -#endif
>> -
>>  #if defined(CONFIG_SPARC) || !defined(CONFIG_OF)
>>  /* Dummy ref counting routines - to be implemented later */
>>  static inline struct device_node *of_node_get(struct device_node *node)
>
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists