Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On Thu, Oct 25, 2018 at 10:53:12AM -0700, Florian Fainelli wrote: > Something like this below? It does not have to be an alternative > solution, I would find it useful for perf to make sure the vectors page > is present in the virtual address space by having an explicit test. perf > maintains, what do you think? Yep, I'd swap the two tests merely because the check for 'sig' will be slightly cheaper than the mask-and-test of addr, but it probably doesn't make that much difference. > > diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c > index f4ea4c62c613..b045b48d368d 100644 > --- a/arch/arm/mm/fault.c > +++ b/arch/arm/mm/fault.c > @@ -173,6 +173,11 @@ __do_user_fault(struct task_struct *tsk, unsigned > long addr, > show_regs(regs); > } > #endif > +#ifndef CONFIG_KUSER_HELPERS > + if (((addr & PAGE_MASK) == 0x) && (sig == SIGSEGV)) > + printk(KERN_DEBUG "%s: CONFIG_KUSER_HELPERS disabled at > 0x%08lx\n", > + tsk->comm, addr); > +#endif > > tsk->thread.address = addr; > tsk->thread.error_code = fsr; > > > > > > BTW, I note that the help text for CONFIG_KUSER_HELPERS needs to be > > updated. > > > > > -- > Florian -- RMK's Patch system: http://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On Thu, Oct 25, 2018 at 10:53:12AM -0700, Florian Fainelli wrote: > Something like this below? It does not have to be an alternative > solution, I would find it useful for perf to make sure the vectors page > is present in the virtual address space by having an explicit test. perf > maintains, what do you think? Yep, I'd swap the two tests merely because the check for 'sig' will be slightly cheaper than the mask-and-test of addr, but it probably doesn't make that much difference. > > diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c > index f4ea4c62c613..b045b48d368d 100644 > --- a/arch/arm/mm/fault.c > +++ b/arch/arm/mm/fault.c > @@ -173,6 +173,11 @@ __do_user_fault(struct task_struct *tsk, unsigned > long addr, > show_regs(regs); > } > #endif > +#ifndef CONFIG_KUSER_HELPERS > + if (((addr & PAGE_MASK) == 0x) && (sig == SIGSEGV)) > + printk(KERN_DEBUG "%s: CONFIG_KUSER_HELPERS disabled at > 0x%08lx\n", > + tsk->comm, addr); > +#endif > > tsk->thread.address = addr; > tsk->thread.error_code = fsr; > > > > > > BTW, I note that the help text for CONFIG_KUSER_HELPERS needs to be > > updated. > > > > > -- > Florian -- RMK's Patch system: http://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On 10/25/18 10:32 AM, Russell King - ARM Linux wrote: > On Thu, Oct 25, 2018 at 10:19:54AM -0700, Florian Fainelli wrote: >> On 10/24/18 7:10 PM, Andrew Lunn wrote: >>> On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some independance with respect to the ARM CPU being used. Add a test which tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is turned on to help asses the system's health. >>> >>> Hi Florian >>> >>> I've suffered the pain from missing CONFIG_KUSER_HELPERS. The >>> segfaults give little clue as to what is going wrong,and gdb is not >>> much use either. >> >> If you have a working backtrace, you can typically see the virtual >> address being in 0x_ which gives a bit of a clue, but yes, this >> is not particularly helpful. >> >>> >>> What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test >>> fails, could you print a message suggesting CONFIG_KUSER_HELPERS on >>> ARM? >> >> Sure, sounds reasonable, thanks for taking a look. > > An alternative would be rather than adding this to user programs, to > have the kernel detect it and print a warning itself. Something like this below? It does not have to be an alternative solution, I would find it useful for perf to make sure the vectors page is present in the virtual address space by having an explicit test. perf maintains, what do you think? diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c index f4ea4c62c613..b045b48d368d 100644 --- a/arch/arm/mm/fault.c +++ b/arch/arm/mm/fault.c @@ -173,6 +173,11 @@ __do_user_fault(struct task_struct *tsk, unsigned long addr, show_regs(regs); } #endif +#ifndef CONFIG_KUSER_HELPERS + if (((addr & PAGE_MASK) == 0x) && (sig == SIGSEGV)) + printk(KERN_DEBUG "%s: CONFIG_KUSER_HELPERS disabled at 0x%08lx\n", + tsk->comm, addr); +#endif tsk->thread.address = addr; tsk->thread.error_code = fsr; > > BTW, I note that the help text for CONFIG_KUSER_HELPERS needs to be > updated. > -- Florian
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On 10/25/18 10:32 AM, Russell King - ARM Linux wrote: > On Thu, Oct 25, 2018 at 10:19:54AM -0700, Florian Fainelli wrote: >> On 10/24/18 7:10 PM, Andrew Lunn wrote: >>> On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some independance with respect to the ARM CPU being used. Add a test which tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is turned on to help asses the system's health. >>> >>> Hi Florian >>> >>> I've suffered the pain from missing CONFIG_KUSER_HELPERS. The >>> segfaults give little clue as to what is going wrong,and gdb is not >>> much use either. >> >> If you have a working backtrace, you can typically see the virtual >> address being in 0x_ which gives a bit of a clue, but yes, this >> is not particularly helpful. >> >>> >>> What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test >>> fails, could you print a message suggesting CONFIG_KUSER_HELPERS on >>> ARM? >> >> Sure, sounds reasonable, thanks for taking a look. > > An alternative would be rather than adding this to user programs, to > have the kernel detect it and print a warning itself. Something like this below? It does not have to be an alternative solution, I would find it useful for perf to make sure the vectors page is present in the virtual address space by having an explicit test. perf maintains, what do you think? diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c index f4ea4c62c613..b045b48d368d 100644 --- a/arch/arm/mm/fault.c +++ b/arch/arm/mm/fault.c @@ -173,6 +173,11 @@ __do_user_fault(struct task_struct *tsk, unsigned long addr, show_regs(regs); } #endif +#ifndef CONFIG_KUSER_HELPERS + if (((addr & PAGE_MASK) == 0x) && (sig == SIGSEGV)) + printk(KERN_DEBUG "%s: CONFIG_KUSER_HELPERS disabled at 0x%08lx\n", + tsk->comm, addr); +#endif tsk->thread.address = addr; tsk->thread.error_code = fsr; > > BTW, I note that the help text for CONFIG_KUSER_HELPERS needs to be > updated. > -- Florian
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On Thu, Oct 25, 2018 at 10:19:54AM -0700, Florian Fainelli wrote: > On 10/24/18 7:10 PM, Andrew Lunn wrote: > > On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: > >> perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some > >> independance with respect to the ARM CPU being used. Add a test which > >> tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is > >> turned on to help asses the system's health. > > > > Hi Florian > > > > I've suffered the pain from missing CONFIG_KUSER_HELPERS. The > > segfaults give little clue as to what is going wrong,and gdb is not > > much use either. > > If you have a working backtrace, you can typically see the virtual > address being in 0x_ which gives a bit of a clue, but yes, this > is not particularly helpful. > > > > > What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test > > fails, could you print a message suggesting CONFIG_KUSER_HELPERS on > > ARM? > > Sure, sounds reasonable, thanks for taking a look. An alternative would be rather than adding this to user programs, to have the kernel detect it and print a warning itself. BTW, I note that the help text for CONFIG_KUSER_HELPERS needs to be updated. -- RMK's Patch system: http://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On Thu, Oct 25, 2018 at 10:19:54AM -0700, Florian Fainelli wrote: > On 10/24/18 7:10 PM, Andrew Lunn wrote: > > On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: > >> perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some > >> independance with respect to the ARM CPU being used. Add a test which > >> tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is > >> turned on to help asses the system's health. > > > > Hi Florian > > > > I've suffered the pain from missing CONFIG_KUSER_HELPERS. The > > segfaults give little clue as to what is going wrong,and gdb is not > > much use either. > > If you have a working backtrace, you can typically see the virtual > address being in 0x_ which gives a bit of a clue, but yes, this > is not particularly helpful. > > > > > What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test > > fails, could you print a message suggesting CONFIG_KUSER_HELPERS on > > ARM? > > Sure, sounds reasonable, thanks for taking a look. An alternative would be rather than adding this to user programs, to have the kernel detect it and print a warning itself. BTW, I note that the help text for CONFIG_KUSER_HELPERS needs to be updated. -- RMK's Patch system: http://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On 10/24/18 7:10 PM, Andrew Lunn wrote: > On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: >> perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some >> independance with respect to the ARM CPU being used. Add a test which >> tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is >> turned on to help asses the system's health. > > Hi Florian > > I've suffered the pain from missing CONFIG_KUSER_HELPERS. The > segfaults give little clue as to what is going wrong,and gdb is not > much use either. If you have a working backtrace, you can typically see the virtual address being in 0x_ which gives a bit of a clue, but yes, this is not particularly helpful. > > What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test > fails, could you print a message suggesting CONFIG_KUSER_HELPERS on > ARM? Sure, sounds reasonable, thanks for taking a look. -- Florian
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On 10/24/18 7:10 PM, Andrew Lunn wrote: > On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: >> perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some >> independance with respect to the ARM CPU being used. Add a test which >> tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is >> turned on to help asses the system's health. > > Hi Florian > > I've suffered the pain from missing CONFIG_KUSER_HELPERS. The > segfaults give little clue as to what is going wrong,and gdb is not > much use either. If you have a working backtrace, you can typically see the virtual address being in 0x_ which gives a bit of a clue, but yes, this is not particularly helpful. > > What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test > fails, could you print a message suggesting CONFIG_KUSER_HELPERS on > ARM? Sure, sounds reasonable, thanks for taking a look. -- Florian
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: > perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some > independance with respect to the ARM CPU being used. Add a test which > tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is > turned on to help asses the system's health. Hi Florian I've suffered the pain from missing CONFIG_KUSER_HELPERS. The segfaults give little clue as to what is going wrong,and gdb is not much use either. What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test fails, could you print a message suggesting CONFIG_KUSER_HELPERS on ARM? Andrew
Re: [PATCH 2/2] perf tests: Add a test for the ARM 32-bit [vectors] page
On Wed, Oct 24, 2018 at 05:09:05PM -0700, Florian Fainelli wrote: > perf on ARM requires CONFIG_KUSER_HELPERS to be turned on to allow some > independance with respect to the ARM CPU being used. Add a test which > tries to locate the [vectors] page, created when CONFIG_KUSER_HELPERS is > turned on to help asses the system's health. Hi Florian I've suffered the pain from missing CONFIG_KUSER_HELPERS. The segfaults give little clue as to what is going wrong,and gdb is not much use either. What i don't see here is any clue to CONFIG_KUSER_HELPERS. If the test fails, could you print a message suggesting CONFIG_KUSER_HELPERS on ARM? Andrew