Based on kernel version 2.6.27. Page generated on 2008-10-13 09:53 EST.
1 Information about /proc/ppc_htab 2 ===================================================================== 3 4 This document and the related code was written by me (Cort Dougan), please 5 email me (cort[AT]fsmlabs.com) if you have questions, comments or corrections[DOT] 6 7 Last Change: 2.16.98 8 9 This entry in the proc directory is readable by all users but only 10 writable by root. 11 12 The ppc_htab interface is a user level way of accessing the 13 performance monitoring registers as well as providing information 14 about the PTE hash table. 15 16 1. Reading 17 18 Reading this file will give you information about the memory management 19 hash table that serves as an extended tlb for page translation on the 20 powerpc. It will also give you information about performance measurement 21 specific to the cpu that you are using. 22 23 Explanation of the 604 Performance Monitoring Fields: 24 MMCR0 - the current value of the MMCR0 register 25 PMC1 26 PMC2 - the value of the performance counters and a 27 description of what events they are counting 28 which are based on MMCR0 bit settings. 29 Explanation of the PTE Hash Table fields: 30 31 Size - hash table size in Kb. 32 Buckets - number of buckets in the table. 33 Address - the virtual kernel address of the hash table base. 34 Entries - the number of ptes that can be stored in the hash table. 35 User/Kernel - how many pte's are in use by the kernel or user at that time. 36 Overflows - How many of the entries are in their secondary hash location. 37 Percent full - ratio of free pte entries to in use entries. 38 Reloads - Count of how many hash table misses have occurred 39 that were fixed with a reload from the linux tables. 40 Should always be 0 on 603 based machines. 41 Non-error Misses - Count of how many hash table misses have occurred 42 that were completed with the creation of a pte in the linux 43 tables with a call to do_page_fault(). 44 Error Misses - Number of misses due to errors such as bad address 45 and permission violations. This includes kernel access of 46 bad user addresses that are fixed up by the trap handler. 47 48 Note that calculation of the data displayed from /proc/ppc_htab takes 49 a long time and spends a great deal of time in the kernel. It would 50 be quite hard on performance to read this file constantly. In time 51 there may be a counter in the kernel that allows successive reads from 52 this file only after a given amount of time has passed to reduce the 53 possibility of a user slowing the system by reading this file. 54 55 2. Writing 56 57 Writing to the ppc_htab allows you to change the characteristics of 58 the powerpc PTE hash table and setup performance monitoring. 59 60 Resizing the PTE hash table is not enabled right now due to many 61 complications with moving the hash table, rehashing the entries 62 and many many SMP issues that would have to be dealt with. 63 64 Write options to ppc_htab: 65 66 - To set the size of the hash table to 64Kb: 67 68 echo 'size 64' > /proc/ppc_htab 69 70 The size must be a multiple of 64 and must be greater than or equal to 71 64. 72 73 - To turn off performance monitoring: 74 75 echo 'off' > /proc/ppc_htab 76 77 - To reset the counters without changing what they're counting: 78 79 echo 'reset' > /proc/ppc_htab 80 81 Note that counting will continue after the reset if it is enabled. 82 83 - To count only events in user mode or only in kernel mode: 84 85 echo 'user' > /proc/ppc_htab 86 ...or... 87 echo 'kernel' > /proc/ppc_htab 88 89 Note that these two options are exclusive of one another and the 90 lack of either of these options counts user and kernel. 91 Using 'reset' and 'off' reset these flags. 92 93 - The 604 has 2 performance counters which can each count events from 94 a specific set of events. These sets are disjoint so it is not 95 possible to count _any_ combination of 2 events. One event can 96 be counted by PMC1 and one by PMC2. 97 98 To start counting a particular event use: 99 100 echo 'event' > /proc/ppc_htab 101 102 and choose from these events: 103 104 PMC1 105 ---- 106 'ic miss' - instruction cache misses 107 'dtlb' - data tlb misses (not hash table misses) 108 109 PMC2 110 ---- 111 'dc miss' - data cache misses 112 'itlb' - instruction tlb misses (not hash table misses) 113 'load miss time' - cycles to complete a load miss 114 115 3. Bugs 116 117 The PMC1 and PMC2 counters can overflow and give no indication of that 118 in /proc/ppc_htab.