Christian Costa
2014-Aug-25 18:58 UTC
[Nouveau] [PATCH envytools] demmio: Add decoding of some MEM_TIMINGS registers for NVC0.
--- rnndb/memory/nvc0_pbfb.xml | 37 ++++++++++++++++++++++++++++++++++--- 1 file changed, 34 insertions(+), 3 deletions(-) diff --git a/rnndb/memory/nvc0_pbfb.xml b/rnndb/memory/nvc0_pbfb.xml index 500cea9..e006dbe 100644 --- a/rnndb/memory/nvc0_pbfb.xml +++ b/rnndb/memory/nvc0_pbfb.xml @@ -49,23 +49,54 @@ Most bitfields are unknown. </doc> <bitfield high="7" low="0" name="RP"> - <doc> RAS# precharge. </doc> + <doc> RAS# Precharge. </doc> </bitfield> <bitfield high="15" low="8" name="RAS"> - <doc>Activate to precharge delay. </doc> + <doc> Activate to precharge delay. </doc> </bitfield> <bitfield high="23" low="16" name="RFC"> <doc> Refresh to activate delay. </doc> </bitfield> <bitfield high="31" low="24" name="RC"> - <doc> Row cycle time. </doc> + <doc> Row Cycle time. </doc> + </bitfield> + </reg32> + + <reg32 offset="0x290" name="MEM_TIMINGS_0" variants="NVC0-"> + <doc> This, and the next 6 regs, are all related to memtimings. + A good place to read might be http://www.tweakers.fr/timings.html . + Most bitfields are unknown. + </doc> + <bitfield high="7" low="0" name="tRC"> + <doc> Row Cycle time. </doc> + </bitfield> + <bitfield high="15" low="8" name="tRFC"> + <doc> Refresh to activate delay. </doc> + </bitfield> + <bitfield high="23" low="17" name="tRAS"> + <doc> Activate to precharge delay. </doc> + </bitfield> + <bitfield high="31" low="24" name="tRCD"> + <doc> RAS# to CAS# Delay. </doc> </bitfield> </reg32> <reg32 offset="0x294" name="MEM_TIMINGS_1" variants="NVC0-"> + <bitfield high="3" low="0" name="tCL"> + <doc> Row Cycle time. </doc> + </bitfield> + <bitfield high="13" low="7" name="tCWL"> + <doc> CAS# Write Latency. </doc> + </bitfield> </reg32> <reg32 offset="0x298" name="MEM_TIMINGS_2" variants="NVC0-"> + <bitfield high="15" low="8" name="tWTR"> + <doc> Write To Read delay. </doc> + </bitfield> + <bitfield high="23" low="15" name="tWR"> + <doc> Write Recovery time. </doc> + </bitfield> </reg32> <reg32 offset="0x29c" name="MEM_TIMINGS_3" variants="NVC0-"> -- 1.9.1
Christian Costa
2014-Aug-25 18:58 UTC
[Nouveau] [PATCH envytools] docs: Fix some spelling and remove some trailing spaces.
--- docs/hw/fifo/intro.rst | 32 ++++++++++++++++---------------- 1 file changed, 16 insertions(+), 16 deletions(-) diff --git a/docs/hw/fifo/intro.rst b/docs/hw/fifo/intro.rst index 8b1d69d..0e023bf 100644 --- a/docs/hw/fifo/intro.rst +++ b/docs/hw/fifo/intro.rst @@ -10,7 +10,7 @@ FIFO overview Introduction =========== -Commands to most of the engines are sent through a special engine caled PFIFO. +Commands to most of the engines are sent through a special engine called PFIFO. PFIFO maintains multiple fully independent command queues, known as "channels" or "FIFO"s. Each channel is controlled through a "channel control area", which is a region of MMIO [pre-NVC0] or VRAM [NVC0+]. PFIFO intercepts all accesses @@ -40,11 +40,11 @@ The engines that PFIFO controls on NV04:NVC0 are: == ========== =========================== =================================================== Id Present on Name Description == ========== =========================== =================================================== -0 all SOFTWARE Not really an engine, causes interrupt for each - command, can be used to execute driver functions - in sync with other commands. -1 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. -2 NV31:NV98 :ref:`PMPEG <pmpeg>` The PFIFO interface to VPE MPEG2 decoding engine. +0 all SOFTWARE Not really an engine, causes interrupt for each + command, can be used to execute driver functions + in sync with other commands. +1 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. +2 NV31:NV98 :ref:`PMPEG <pmpeg>` The PFIFO interface to VPE MPEG2 decoding engine. NVA0:NVAA 3 NV40:NV84 :ref:`PME <me-fifo>` VPE motion estimation engine. 4 NV41:NV84 :ref:`PVP1 <pvp1>` VPE microcoded vector processor. @@ -64,10 +64,10 @@ The engines that PFIFO controls on NVC0- are: ===== ========== =========================== =================================================== Id Present on Name Description ===== ========== =========================== =================================================== -1f all SOFTWARE Not really an engine, causes interrupt for each - command, can be used to execute driver functions - in sync with other commands. -0 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. +1f all SOFTWARE Not really an engine, causes interrupt for each + command, can be used to execute driver functions + in sync with other commands. +0 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. 1 all :ref:`PVDEC <pvdec>` falcon-based microcoded video decoder. 2 all :ref:`PPPP <pppp>` falcon-based video post-processor. 3 all :ref:`PVLD <pvld>` falcon-based variable length decoder. @@ -93,7 +93,7 @@ The PFIFO can be split into roughly 4 pieces: - PFIFO puller: executes the commands, passes them to the proper engine, or to the driver. - PFIFO switcher: ticks out the time slices for the channels and saves / - restores the state of the channel between PFIFO registers and RAMFC + restores the state of the channels between PFIFO registers and RAMFC memory. A channel consists of the following: @@ -133,7 +133,7 @@ is being submitted. See :ref:`nv03-pfifo-dma` for details. NV04 PFIFO greatly enhanced the DMA mode and made it controllable directly through the channel control area. Thus, commands can now be submitted by -multiple applications simultanously, without coordination with each other +multiple applications simultaneously, without coordination with each other and without kernel's help. DMA mode is described in :ref:`fifo-dma-pusher`. NV50 introduced IB mode. IB mode is a modified version of DMA mode that, @@ -143,8 +143,8 @@ to stitch together parts of multiple memory areas into a single command stream memory written by earlier commands. IB mode is described along with DMA mode in :ref:`fifo-dma-pusher`. -NVC0 rearchitected the whole PFIFO, made it possible to have up to 3 channels -executing simultanously, and introduced a new DMA packet format. +NVC0 rearchitectured the whole PFIFO, made it possible to have up to 3 channels +executing simultaneously, and introduced a new DMA packet format. The commands, as stored in CACHE, are tuples of: @@ -180,12 +180,12 @@ and engine-specific documentation. Pre-NV1A, PFIFO treats everything as little-endian. NV1A introduced big-endian mode, which affects pushbuffer/IB reads and semaphores. On NV1A:NV50 cards, -the endian can be selected per channel via big_endian flag. On NV50+ cards, +the endianness can be selected per channel via the big_endian flag. On NV50+ cards, PFIFO endianness is a global switch. .. todo:: look for NVC0 PFIFO endian switch The channel control area endianness is not affected by the big_endian flag or -NV50+ PFIFO endianness switch. Instead, it follows the PMC MMIO endian switch. +NV50+ PFIFO endianness switch. Instead, it follows the PMC MMIO endianness switch. .. todo:: is it still true for NVC0, with VRAM-backed channel control area? -- 1.9.1
Christian Costa
2014-Aug-25 18:58 UTC
[Nouveau] [PATCH envytools] nva: Clean up nva tools doc
- Sort commands by name for easier update - Make more readable - Remove no more existing commands --- nva/README | 83 +++++++++++++++++++++++++++++++++----------------------------- 1 file changed, 44 insertions(+), 39 deletions(-) diff --git a/nva/README b/nva/README index 625ea7c..d4347c5 100644 --- a/nva/README +++ b/nva/README @@ -3,7 +3,7 @@ registers. WARNING: these tools *can* and *will* hang your machine if you don't know what you're doing. Hardware destruction is likely also possible, although -no incidents are known to have happened yet. In most cases it's also not +no incident are known to have happened yet. In most cases it's also not recommended to use these tools while a driver is active for a given card. All programs except nvalist take an optional -c <card number> parameter. @@ -13,43 +13,48 @@ by the nvalist program. The programs are: -nvalist: prints a list of cards - -nvapeek <address> [<byte count>]: reads 32-bit MMIO register at <address>. -If byte count is also given, reads all registers in range [address, address -+ byte count). - -nvapeek8 <address> [<byte count>]: like nvapeek, but does 8-bit MMIO accesses - -nvapoke <address> <value>: writes a 32-bit <value> to the MMIO register at -<address>. - -nvapoke8 <address> <value>: like nvapoke, but does 8-bit MMIO access - -nvawatch [-t] <address>: reads MMIO register at <address> in a loop, prints -the value every time it changes. If -t is specified, prints a timestamp -and diff from the previous timestamp before the value. Never quits, needs -to be manually aborted. - -nvahammer <address> <value>: like nvapoke, but repeats the write in -an infinite loop. Needs to be manually aborted. - -nvagetbios [-s <extraction method>]: extracts the card's VBIOS using the -method given as parameter and writes it to stdout. Method can be PROM or -PRAMIN. If method is not given, defaults to something sensible. - -nvafuzz <address> [<byte count>]: writes random values to a register or -a register range in an infinite loop. Needs to be manually aborted. - -nvascan [-as] <address> [<byte count>]: for each register in a range: -read it, write 0xffffffff, read it, write 0, read it, write back the -original value. Helpful to see the valid values for registers. If -s option -is passed, does a slow scan - waits and reads PMC.ID register between scans -to recover from errors caused by invalid register accesses. If -a option -is passed, does a cross-test on all registers in the range to detect aliased -addresses [not particularly reliable]. - -nvatiming: attempts to measure what frequency various units of the card are -running at by using misc tricks. +nvafuzz <address> [<byte count>] + Writes random values to a register or + a register range in an infinite loop. Needs to be manually aborted. + +nvagetbios [-s <extraction method>] + Extracts the card's VBIOS using the + method given as parameter and writes it to stdout. Method can be PROM or + PRAMIN. If method is not given, defaults to something sensible. + +nvahammer <address> <value> + Like nvapoke, but repeats the write in + an infinite loop. Needs to be manually aborted. + +nvalist + Prints a list of cards + +nvapeek <address> [<byte count>] + Reads 32-bit MMIO register at <address>. + If byte count is also given, reads all registers in range [address, address + + byte count). + +nvapoke <address> <value> + Writes a 32-bit <value> to the MMIO register at + <address>. + +nvascan [-as] <address> [<byte count>] + For each register in a range: + read it, write 0xffffffff, read it, write 0, read it, write back the + original value. Helpful to see the valid values for registers. If -s option + is passed, does a slow scan - waits and reads PMC.ID register between scans + to recover from errors caused by invalid register accesses. If -a option + is passed, does a cross-test on all registers in the range to detect aliased + addresses [not particularly reliable]. + +nvatiming + Attempts to measure what frequency various units of the card are + running at by using misc tricks. + +nvawatch [-t] <address> + Reads MMIO register at <address> in a loop, prints + the value every time it changes. If -t is specified, prints a timestamp + and diff from the previous timestamp before the value. Never quits, needs + to be manually aborted. [XXX: document the remaining tools] -- 1.9.1
Christian Costa
2014-Aug-25 18:58 UTC
[Nouveau] [PATCH envytools] nvbios: Return an error if mem train subentries is above the limit to avoid stack corruption.
--- nvbios/mem.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/nvbios/mem.c b/nvbios/mem.c index 911ce5c..2c8a929 100644 --- a/nvbios/mem.c +++ b/nvbios/mem.c @@ -111,6 +111,10 @@ envy_bios_parse_mem_train (struct envy_bios *bios) { struct envy_bios_mem_train_entry *entry = &mt->entries[i]; entry->offset = mt->offset + mt->hlen + ((mt->rlen + mt->subentries * mt->subentrylen) * i); err |= bios_u8(bios, entry->offset, &entry->u00); + if (mt->subentries > sizeof(entry->subentry)) { + ENVY_BIOS_ERR("Error when parsing mem train: subentries = %d > %lu\n", mt->subentries, sizeof(entry->subentry)); + return -EFAULT; + } int j; for (j = 0; j < mt->subentries; j++) { err |= bios_u8(bios, entry->offset+j+1, &entry->subentry[j]); -- 1.9.1
Christian Costa
2014-Aug-25 18:58 UTC
[Nouveau] [PATCH envytools] nvbios: Write missing null terminating byte.
--- nvbios/nvbios.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/nvbios/nvbios.c b/nvbios/nvbios.c index 4fc667b..7bb768a 100644 --- a/nvbios/nvbios.c +++ b/nvbios/nvbios.c @@ -797,9 +797,9 @@ void find_strap(char *filename) { path = (char*) malloc(base_length + strlen(strap_filename)+1); strncpy(path, filename, base_length); - strncpy(path+base_length, strap_filename, strlen(strap_filename)); + strncpy(path+base_length, strap_filename, strlen(strap_filename)+1); - if(!set_strap_from_file(path)) + if (!set_strap_from_file(path)) printf("Strap register found in '%s'\n", path); free(path); -- 1.9.1
Christian Costa
2014-Aug-25 18:58 UTC
[Nouveau] [PATCH envytools] nvbios: Check if HWEA entry length == -1.
--- nvbios/bios.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/nvbios/bios.c b/nvbios/bios.c index fac3840..ca8c54e 100644 --- a/nvbios/bios.c +++ b/nvbios/bios.c @@ -276,9 +276,14 @@ int envy_bios_parse (struct envy_bios *bios) { pos += 4; if (entry.type > 1) ENVY_BIOS_WARN("Unknown HWEA entry type %d\n", entry.type); + if (entry.len == (uint32_t)-1) { + /* Maybe this should be treated as for length == 0 */ + ENVY_BIOS_ERR("Invalid HWEA entry length %d\n", entry.len); + break; + } } else { entry.base = word & 0x1fffffc; - entry.type = word&3; + entry.type = word & 3; switch (entry.type) { case 0: case 1: -- 1.9.1
Christian Costa
2014-Aug-25 18:58 UTC
[Nouveau] [PATCH envytools] Don't try to use bios->pars if NULL (parse_pcir function has probably failed).
--- nvbios/info.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/nvbios/info.c b/nvbios/info.c index 87a33d8..6b8d209 100644 --- a/nvbios/info.c +++ b/nvbios/info.c @@ -119,6 +119,10 @@ int envy_bios_parse_bit_i (struct envy_bios *bios, struct envy_bios_bit_entry *b bios->chipset_name = "G72"; break; case 0x0573: + if (!bios->parts) { + ENVY_BIOS_ERR("Failed to detect board (bios->pars is NULL, parse_pcir function has probably failed)\n"); + break; + } /* yeah, ain't this one hilarious... */ if ((bios->parts[0].pcir_device & 0xfff0) == 0x7e0) { bios->chipset = 0x63; -- 1.9.1
Ilia Mirkin
2014-Aug-25 19:09 UTC
[Nouveau] [PATCH envytools] Don't try to use bios->pars if NULL (parse_pcir function has probably failed).
On Mon, Aug 25, 2014 at 2:58 PM, Christian Costa <titan.costa at gmail.com> wrote:> --- > nvbios/info.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/nvbios/info.c b/nvbios/info.c > index 87a33d8..6b8d209 100644 > --- a/nvbios/info.c > +++ b/nvbios/info.c > @@ -119,6 +119,10 @@ int envy_bios_parse_bit_i (struct envy_bios *bios, struct envy_bios_bit_entry *b > bios->chipset_name = "G72"; > break; > case 0x0573: > + if (!bios->parts) { > + ENVY_BIOS_ERR("Failed to detect board (bios->pars is NULL, parse_pcir function has probably failed)\n");parts> + break; > + } > /* yeah, ain't this one hilarious... */ > if ((bios->parts[0].pcir_device & 0xfff0) == 0x7e0) { > bios->chipset = 0x63; > -- > 1.9.1 > > _______________________________________________ > Nouveau mailing list > Nouveau at lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/nouveau
Martin Peres
2014-Aug-25 22:30 UTC
[Nouveau] [PATCH envytools] demmio: Add decoding of some MEM_TIMINGS registers for NVC0.
On 25/08/2014 20:58, Christian Costa wrote:> --- > rnndb/memory/nvc0_pbfb.xml | 37 ++++++++++++++++++++++++++++++++++--- > 1 file changed, 34 insertions(+), 3 deletions(-) > > diff --git a/rnndb/memory/nvc0_pbfb.xml b/rnndb/memory/nvc0_pbfb.xml > index 500cea9..e006dbe 100644 > --- a/rnndb/memory/nvc0_pbfb.xml > +++ b/rnndb/memory/nvc0_pbfb.xml > @@ -49,23 +49,54 @@ > Most bitfields are unknown. > </doc> > <bitfield high="7" low="0" name="RP"> > - <doc> RAS# precharge. </doc> > + <doc> RAS# Precharge. </doc> > </bitfield> > <bitfield high="15" low="8" name="RAS"> > - <doc>Activate to precharge delay. </doc> > + <doc> Activate to precharge delay. </doc> > </bitfield> > <bitfield high="23" low="16" name="RFC"> > <doc> Refresh to activate delay. </doc> > </bitfield> > <bitfield high="31" low="24" name="RC"> > - <doc> Row cycle time. </doc> > + <doc> Row Cycle time. </doc> > + </bitfield> > + </reg32> > + > + <reg32 offset="0x290" name="MEM_TIMINGS_0" variants="NVC0-"> > + <doc> This, and the next 6 regs, are all related to memtimings. > + A good place to read might be http://www.tweakers.fr/timings.html . > + Most bitfields are unknown. > + </doc> > + <bitfield high="7" low="0" name="tRC"> > + <doc> Row Cycle time. </doc> > + </bitfield> > + <bitfield high="15" low="8" name="tRFC"> > + <doc> Refresh to activate delay. </doc> > + </bitfield> > + <bitfield high="23" low="17" name="tRAS"> > + <doc> Activate to precharge delay. </doc> > + </bitfield> > + <bitfield high="31" low="24" name="tRCD"> > + <doc> RAS# to CAS# Delay. </doc> > </bitfield> > </reg32> > > <reg32 offset="0x294" name="MEM_TIMINGS_1" variants="NVC0-"> > + <bitfield high="3" low="0" name="tCL"> > + <doc> Row Cycle time. </doc> > + </bitfield> > + <bitfield high="13" low="7" name="tCWL"> > + <doc> CAS# Write Latency. </doc> > + </bitfield> > </reg32> > > <reg32 offset="0x298" name="MEM_TIMINGS_2" variants="NVC0-"> > + <bitfield high="15" low="8" name="tWTR"> > + <doc> Write To Read delay. </doc> > + </bitfield> > + <bitfield high="23" low="15" name="tWR"> > + <doc> Write Recovery time. </doc> > + </bitfield> > </reg32> > > <reg32 offset="0x29c" name="MEM_TIMINGS_3" variants="NVC0-">Thanks for the patch but where did you get this knowledge? By comparing with nv50 and the vbios table?
Martin Peres
2014-Aug-25 22:32 UTC
[Nouveau] [PATCH envytools] docs: Fix some spelling and remove some trailing spaces.
On 25/08/2014 20:58, Christian Costa wrote:> --- > docs/hw/fifo/intro.rst | 32 ++++++++++++++++---------------- > 1 file changed, 16 insertions(+), 16 deletions(-) > > diff --git a/docs/hw/fifo/intro.rst b/docs/hw/fifo/intro.rst > index 8b1d69d..0e023bf 100644 > --- a/docs/hw/fifo/intro.rst > +++ b/docs/hw/fifo/intro.rst > @@ -10,7 +10,7 @@ FIFO overview > Introduction > ===========> > -Commands to most of the engines are sent through a special engine caled PFIFO. > +Commands to most of the engines are sent through a special engine called PFIFO. > PFIFO maintains multiple fully independent command queues, known as "channels" > or "FIFO"s. Each channel is controlled through a "channel control area", which > is a region of MMIO [pre-NVC0] or VRAM [NVC0+]. PFIFO intercepts all accesses > @@ -40,11 +40,11 @@ The engines that PFIFO controls on NV04:NVC0 are: > == ========== =========================== ==================================================> Id Present on Name Description > == ========== =========================== ==================================================> -0 all SOFTWARE Not really an engine, causes interrupt for each > - command, can be used to execute driver functions > - in sync with other commands. > -1 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. > -2 NV31:NV98 :ref:`PMPEG <pmpeg>` The PFIFO interface to VPE MPEG2 decoding engine. > +0 all SOFTWARE Not really an engine, causes interrupt for each > + command, can be used to execute driver functions > + in sync with other commands. > +1 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. > +2 NV31:NV98 :ref:`PMPEG <pmpeg>` The PFIFO interface to VPE MPEG2 decoding engine. > NVA0:NVAA > 3 NV40:NV84 :ref:`PME <me-fifo>` VPE motion estimation engine. > 4 NV41:NV84 :ref:`PVP1 <pvp1>` VPE microcoded vector processor. > @@ -64,10 +64,10 @@ The engines that PFIFO controls on NVC0- are: > ===== ========== =========================== ==================================================> Id Present on Name Description > ===== ========== =========================== ==================================================> -1f all SOFTWARE Not really an engine, causes interrupt for each > - command, can be used to execute driver functions > - in sync with other commands. > -0 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. > +1f all SOFTWARE Not really an engine, causes interrupt for each > + command, can be used to execute driver functions > + in sync with other commands. > +0 all :ref:`PGRAPH <graph-intro>` Main engine of the card: 2d, 3d, compute. > 1 all :ref:`PVDEC <pvdec>` falcon-based microcoded video decoder. > 2 all :ref:`PPPP <pppp>` falcon-based video post-processor. > 3 all :ref:`PVLD <pvld>` falcon-based variable length decoder. > @@ -93,7 +93,7 @@ The PFIFO can be split into roughly 4 pieces: > - PFIFO puller: executes the commands, passes them to the proper engine, > or to the driver. > - PFIFO switcher: ticks out the time slices for the channels and saves / > - restores the state of the channel between PFIFO registers and RAMFC > + restores the state of the channels between PFIFO registers and RAMFC > memory. > > A channel consists of the following: > @@ -133,7 +133,7 @@ is being submitted. See :ref:`nv03-pfifo-dma` for details. > > NV04 PFIFO greatly enhanced the DMA mode and made it controllable directly > through the channel control area. Thus, commands can now be submitted by > -multiple applications simultanously, without coordination with each other > +multiple applications simultaneously, without coordination with each other > and without kernel's help. DMA mode is described in :ref:`fifo-dma-pusher`. > > NV50 introduced IB mode. IB mode is a modified version of DMA mode that, > @@ -143,8 +143,8 @@ to stitch together parts of multiple memory areas into a single command stream > memory written by earlier commands. IB mode is described along with DMA mode in > :ref:`fifo-dma-pusher`. > > -NVC0 rearchitected the whole PFIFO, made it possible to have up to 3 channels > -executing simultanously, and introduced a new DMA packet format. > +NVC0 rearchitectured the whole PFIFO, made it possible to have up to 3 channels > +executing simultaneously, and introduced a new DMA packet format. > > The commands, as stored in CACHE, are tuples of: > > @@ -180,12 +180,12 @@ and engine-specific documentation. > > Pre-NV1A, PFIFO treats everything as little-endian. NV1A introduced big-endian > mode, which affects pushbuffer/IB reads and semaphores. On NV1A:NV50 cards, > -the endian can be selected per channel via big_endian flag. On NV50+ cards, > +the endianness can be selected per channel via the big_endian flag. On NV50+ cards, > PFIFO endianness is a global switch. > > .. todo:: look for NVC0 PFIFO endian switch > > The channel control area endianness is not affected by the big_endian flag or > -NV50+ PFIFO endianness switch. Instead, it follows the PMC MMIO endian switch. > +NV50+ PFIFO endianness switch. Instead, it follows the PMC MMIO endianness switch. > > .. todo:: is it still true for NVC0, with VRAM-backed channel control area?Looks good, I'll push it soon!
Martin Peres
2014-Aug-25 22:35 UTC
[Nouveau] [PATCH envytools] nva: Clean up nva tools doc
On 25/08/2014 20:58, Christian Costa wrote:> - Sort commands by name for easier update > - Make more readable > - Remove no more existing commands > --- > nva/README | 83 +++++++++++++++++++++++++++++++++----------------------------- > 1 file changed, 44 insertions(+), 39 deletions(-) > > diff --git a/nva/README b/nva/README > index 625ea7c..d4347c5 100644 > --- a/nva/README > +++ b/nva/README > @@ -3,7 +3,7 @@ registers. > > WARNING: these tools *can* and *will* hang your machine if you don't know > what you're doing. Hardware destruction is likely also possible, although > -no incidents are known to have happened yet. In most cases it's also not > +no incident are known to have happened yet. In most cases it's also notIIRC, "0" is considered plural in English. I'm sure Ilia or Ben could give us a more definitive answer on this :)> recommended to use these tools while a driver is active for a given card. > > All programs except nvalist take an optional -c <card number> parameter. > @@ -13,43 +13,48 @@ by the nvalist program. > > The programs are: > > -nvalist: prints a list of cards > - > -nvapeek <address> [<byte count>]: reads 32-bit MMIO register at <address>. > -If byte count is also given, reads all registers in range [address, address > -+ byte count). > - > -nvapeek8 <address> [<byte count>]: like nvapeek, but does 8-bit MMIO accesses > - > -nvapoke <address> <value>: writes a 32-bit <value> to the MMIO register at > -<address>. > - > -nvapoke8 <address> <value>: like nvapoke, but does 8-bit MMIO access > - > -nvawatch [-t] <address>: reads MMIO register at <address> in a loop, prints > -the value every time it changes. If -t is specified, prints a timestamp > -and diff from the previous timestamp before the value. Never quits, needs > -to be manually aborted. > - > -nvahammer <address> <value>: like nvapoke, but repeats the write in > -an infinite loop. Needs to be manually aborted. > - > -nvagetbios [-s <extraction method>]: extracts the card's VBIOS using the > -method given as parameter and writes it to stdout. Method can be PROM or > -PRAMIN. If method is not given, defaults to something sensible. > - > -nvafuzz <address> [<byte count>]: writes random values to a register or > -a register range in an infinite loop. Needs to be manually aborted. > - > -nvascan [-as] <address> [<byte count>]: for each register in a range: > -read it, write 0xffffffff, read it, write 0, read it, write back the > -original value. Helpful to see the valid values for registers. If -s option > -is passed, does a slow scan - waits and reads PMC.ID register between scans > -to recover from errors caused by invalid register accesses. If -a option > -is passed, does a cross-test on all registers in the range to detect aliased > -addresses [not particularly reliable]. > - > -nvatiming: attempts to measure what frequency various units of the card are > -running at by using misc tricks. > +nvafuzz <address> [<byte count>] > + Writes random values to a register or > + a register range in an infinite loop. Needs to be manually aborted. > + > +nvagetbios [-s <extraction method>] > + Extracts the card's VBIOS using the > + method given as parameter and writes it to stdout. Method can be PROM or > + PRAMIN. If method is not given, defaults to something sensible. > + > +nvahammer <address> <value> > + Like nvapoke, but repeats the write in > + an infinite loop. Needs to be manually aborted. > + > +nvalist > + Prints a list of cards > + > +nvapeek <address> [<byte count>] > + Reads 32-bit MMIO register at <address>. > + If byte count is also given, reads all registers in range [address, address > + + byte count). > + > +nvapoke <address> <value> > + Writes a 32-bit <value> to the MMIO register at > + <address>. > + > +nvascan [-as] <address> [<byte count>] > + For each register in a range: > + read it, write 0xffffffff, read it, write 0, read it, write back the > + original value. Helpful to see the valid values for registers. If -s option > + is passed, does a slow scan - waits and reads PMC.ID register between scans > + to recover from errors caused by invalid register accesses. If -a option > + is passed, does a cross-test on all registers in the range to detect aliased > + addresses [not particularly reliable]. > + > +nvatiming > + Attempts to measure what frequency various units of the card are > + running at by using misc tricks. > + > +nvawatch [-t] <address> > + Reads MMIO register at <address> in a loop, prints > + the value every time it changes. If -t is specified, prints a timestamp > + and diff from the previous timestamp before the value. Never quits, needs > + to be manually aborted. > > [XXX: document the remaining tools]Did you fix some spelling mistakes in while changing the order? I already suck at comparing lines that are one on top of the other, so I don't even want to try checking!
Martin Peres
2014-Aug-25 22:37 UTC
[Nouveau] [PATCH envytools] nvbios: Write missing null terminating byte.
On 25/08/2014 20:58, Christian Costa wrote:> --- > nvbios/nvbios.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/nvbios/nvbios.c b/nvbios/nvbios.c > index 4fc667b..7bb768a 100644 > --- a/nvbios/nvbios.c > +++ b/nvbios/nvbios.c > @@ -797,9 +797,9 @@ void find_strap(char *filename) { > > path = (char*) malloc(base_length + strlen(strap_filename)+1); > strncpy(path, filename, base_length); > - strncpy(path+base_length, strap_filename, strlen(strap_filename)); > + strncpy(path+base_length, strap_filename, strlen(strap_filename)+1); > > - if(!set_strap_from_file(path)) > + if (!set_strap_from_file(path)) > printf("Strap register found in '%s'\n", path); > > free(path);Looks good
Christian Costa
2014-Aug-26 02:06 UTC
[Nouveau] [PATCH envytools] demmio: Add decoding of some MEM_TIMINGS registers for NVC0.
Le 26/08/2014 00:30, Martin Peres a ?crit :> On 25/08/2014 20:58, Christian Costa wrote: >> --- >> rnndb/memory/nvc0_pbfb.xml | 37 ++++++++++++++++++++++++++++++++++--- >> 1 file changed, 34 insertions(+), 3 deletions(-) >> >> diff --git a/rnndb/memory/nvc0_pbfb.xml b/rnndb/memory/nvc0_pbfb.xml >> index 500cea9..e006dbe 100644 >> --- a/rnndb/memory/nvc0_pbfb.xml >> +++ b/rnndb/memory/nvc0_pbfb.xml >> @@ -49,23 +49,54 @@ >> Most bitfields are unknown. >> </doc> >> <bitfield high="7" low="0" name="RP"> >> - <doc> RAS# precharge. </doc> >> + <doc> RAS# Precharge. </doc> >> </bitfield> >> <bitfield high="15" low="8" name="RAS"> >> - <doc>Activate to precharge delay. </doc> >> + <doc> Activate to precharge delay. </doc> >> </bitfield> >> <bitfield high="23" low="16" name="RFC"> >> <doc> Refresh to activate delay. </doc> >> </bitfield> >> <bitfield high="31" low="24" name="RC"> >> - <doc> Row cycle time. </doc> >> + <doc> Row Cycle time. </doc> >> + </bitfield> >> + </reg32> >> + >> + <reg32 offset="0x290" name="MEM_TIMINGS_0" variants="NVC0-"> >> + <doc> This, and the next 6 regs, are all related to memtimings. >> + A good place to read might be >> http://www.tweakers.fr/timings.html . >> + Most bitfields are unknown. >> + </doc> >> + <bitfield high="7" low="0" name="tRC"> >> + <doc> Row Cycle time. </doc> >> + </bitfield> >> + <bitfield high="15" low="8" name="tRFC"> >> + <doc> Refresh to activate delay. </doc> >> + </bitfield> >> + <bitfield high="23" low="17" name="tRAS"> >> + <doc> Activate to precharge delay. </doc> >> + </bitfield> >> + <bitfield high="31" low="24" name="tRCD"> >> + <doc> RAS# to CAS# Delay. </doc> >> </bitfield> >> </reg32> >> <reg32 offset="0x294" name="MEM_TIMINGS_1" variants="NVC0-"> >> + <bitfield high="3" low="0" name="tCL"> >> + <doc> Row Cycle time. </doc> >> + </bitfield> >> + <bitfield high="13" low="7" name="tCWL"> >> + <doc> CAS# Write Latency. </doc> >> + </bitfield> >> </reg32> >> <reg32 offset="0x298" name="MEM_TIMINGS_2" variants="NVC0-"> >> + <bitfield high="15" low="8" name="tWTR"> >> + <doc> Write To Read delay. </doc> >> + </bitfield> >> + <bitfield high="23" low="15" name="tWR"> >> + <doc> Write Recovery time. </doc> >> + </bitfield> >> </reg32> >> <reg32 offset="0x29c" name="MEM_TIMINGS_3" variants="NVC0-"> > > Thanks for the patch but where did you get this knowledge? By > comparing with nv50 and the vbios table?It's based on the code in nvbios.c: reg_100220 = (tRCD << 24 | (tRAS&0x7f) << 17 | tRFC << 8 | tRC); reg_100224 = 0x4c << 24 | (tUNK_11&0x0f) << 20 | (tCWL << 7) | (tCL & 0x0f); reg_100228 = 0x44000011 | tWR << 16 | tWTR << 8; reg_10022c = tUNK_20 << 9 | tUNK_13; reg_100230 = 0x42e00069 | tUNK_12 << 15; printf("Registers: 290: 0x%08x 0x%08x 0x%08x 0x%08x\n", reg_100220, reg_100224, reg_100228, reg_10022c); printf(" 2a0: 0x%08x 0x%08x 0x%08x 0x%08x\n", reg_100230, reg_100234, reg_100238, reg_10023c);
Seemingly Similar Threads
- [PATCH envytools] Don't try to use bios->pars if NULL (parse_pcir function has probably failed).
- [PATCH envytools] nva: Clean up nva tools doc
- [PATCH envytools] demmio: Add decoding of some MEM_TIMINGS registers for NVC0.
- [PATCH envytools] nvbios: Fix reading of ram_restrict_group_count.
- help with ancestral.pars in phangorn package