Messages in this thread | | | Date | Wed, 20 Jun 2007 12:49:05 -0700 (PDT) | From | David Rientjes <> | Subject | Re: [patch 2/8] 2.6.22-rc3 perfmon2 : Debug messages added |
| |
On Wed, 20 Jun 2007, Robert Richter wrote:
> Debug messages added for better debugging. >
And you added BUG_ON()'s.
> Signed-off-by: Robert Richter <robert.richter@amd.com> > > Index: linux-2.6.22-rc3/perfmon/perfmon_file.c > =================================================================== > --- linux-2.6.22-rc3.orig/perfmon/perfmon_file.c > +++ linux-2.6.22-rc3/perfmon/perfmon_file.c > @@ -192,6 +192,8 @@ static int pfm_mmap(struct file *file, s > unsigned long flags; > int ret; > > + PFM_DBG("pfm_file_ops");
After commenting on your first set of patches, I've been using it a little more. In my use, these debugging messages weren't very helpful because "pfm_file_ops" can indicate pfm_mmap, pfm_read, pfm_poll, etc. Could these be changed to be more specific based on the function they're in?
> Index: linux-2.6.22-rc3/perfmon/perfmon_syscalls.c > =================================================================== > --- linux-2.6.22-rc3.orig/perfmon/perfmon_syscalls.c > +++ linux-2.6.22-rc3/perfmon/perfmon_syscalls.c > @@ -403,6 +403,8 @@ asmlinkage long sys_pfm_create_context(s > void *fmt_arg = NULL; > int ret; > > + PFM_DBG("syscall");
Likewise. Using "syscall" for all debugging messages in the syscall handlers isn't very informative. Could this be PFM_DBG(__FUNCTION__) instead?
> + > if (atomic_read(&perfmon_disabled)) > return -ENOSYS; > > @@ -433,8 +435,12 @@ asmlinkage long sys_pfm_write_pmcs(int f > size_t sz; > int ret, fput_needed; > > - if (count < 0 || count >= PFM_MAX_ARG_COUNT(ureq)) > + PFM_DBG("syscall"); > + > + if (count < 0 || count >= PFM_MAX_ARG_COUNT(ureq)) { > + PFM_DBG("invalid arg count %d", count);
This is whitespace damaged.
> return -EINVAL; > + } > > sz = count*sizeof(*ureq); > > @@ -475,6 +481,8 @@ asmlinkage long sys_pfm_write_pmcs(int f > kfree(fptr); > error: > fput_light(filp, fput_needed); > + if (ret) > + PFM_DBG("failed: errno=%d", -ret);
What failed? More information would be helpful since this is, after all, a diagnostic message.
> Index: linux-2.6.22-rc3/include/asm-i386/perfmon.h > =================================================================== > --- linux-2.6.22-rc3.orig/include/asm-i386/perfmon.h > +++ linux-2.6.22-rc3/include/asm-i386/perfmon.h > @@ -140,6 +140,10 @@ static inline void pfm_arch_write_pmc(st > if (ctx && ctx->flags.started == 0) > return; > > + PFM_DBG_ovfl("pfm_arch_write_pmc(0x%016Lx, 0x%016Lx)", > + (unsigned long long) pfm_pmu_conf->pmc_desc[cnum].hw_addr, > + (unsigned long long) value);
Casting here should be unnecessary. Changing %L would be advisible to display the data as it is stored in the object.
> + BUG_ON(pfm_pmu_conf->pmc_desc[cnum].type & PFM_REG_V); > if (arch_info->pmu_style == PFM_X86_PMU_P4) > __pfm_write_reg_p4(&arch_info->pmc_addrs[cnum], value); > else > @@ -155,6 +159,10 @@ static inline void pfm_arch_write_pmd(st > if (pfm_pmu_conf->pmd_desc[cnum].type & PFM_REG_C64) > value |= ~pfm_pmu_conf->ovfl_mask; > > + PFM_DBG_ovfl("pfm_arch_write_pmd(0x%016Lx, 0x%016Lx)", > + (unsigned long long) pfm_pmu_conf->pmd_desc[cnum].hw_addr, > + (unsigned long long) value); > + BUG_ON(pfm_pmu_conf->pmd_desc[cnum].type & PFM_REG_V); > if (arch_info->pmu_style == PFM_X86_PMU_P4) > __pfm_write_reg_p4(&arch_info->pmd_addrs[cnum], value); > else > @@ -165,10 +173,14 @@ static inline u64 pfm_arch_read_pmd(stru > { > struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; > u64 tmp; > + BUG_ON(pfm_pmu_conf->pmd_desc[cnum].type & PFM_REG_V); > if (arch_info->pmu_style == PFM_X86_PMU_P4) > __pfm_read_reg_p4(&arch_info->pmd_addrs[cnum], &tmp); > else > rdmsrl(pfm_pmu_conf->pmd_desc[cnum].hw_addr, tmp); > + PFM_DBG_ovfl("pfm_arch_read_pmd(0x%016Lx) = 0x%016Lx", > + (unsigned long long) pfm_pmu_conf->pmd_desc[cnum].hw_addr, > + (unsigned long long) tmp); > return tmp; > } > > @@ -176,10 +188,14 @@ static inline u64 pfm_arch_read_pmc(stru > { > struct pfm_arch_pmu_info *arch_info = pfm_pmu_conf->arch_info; > u64 tmp; > + BUG_ON(pfm_pmu_conf->pmc_desc[cnum].type & PFM_REG_V); > if (arch_info->pmu_style == PFM_X86_PMU_P4) > __pfm_read_reg_p4(&arch_info->pmc_addrs[cnum], &tmp); > else > rdmsrl(pfm_pmu_conf->pmc_desc[cnum].hw_addr, tmp); > + PFM_DBG_ovfl("pfm_arch_read_pmc(0x%016Lx) = 0x%016Lx", > + (unsigned long long) pfm_pmu_conf->pmc_desc[cnum].hw_addr, > + (unsigned long long) tmp);
More whitespace damage. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |