From: Marcin Wojtas <mw@semihalf.com>
To: Leif Lindholm <leif.lindholm@linaro.org>
Cc: edk2-devel-01 <edk2-devel@lists.01.org>,
Ard Biesheuvel <ard.biesheuvel@linaro.org>,
nadavh@marvell.com, Neta Zur Hershkovits <neta@marvell.com>,
Kostya Porotchkin <kostap@marvell.com>,
Hua Jing <jinghua@marvell.com>,
semihalf-dabros-jan <jsd@semihalf.com>
Subject: Re: [platforms: PATCH 6/8] Marvell/Armada: Enable dynamic DRAM size detection
Date: Thu, 12 Oct 2017 07:47:52 +0200 [thread overview]
Message-ID: <CAPv3WKepp4piZonQLk9mOgg6xa5nJ-Lib76M=ZRnfDv-RRJ0ww@mail.gmail.com> (raw)
In-Reply-To: <20171011175613.ozdhj5n7a6n247qi@bivouac.eciton.net>
2017-10-11 19:56 GMT+02:00 Leif Lindholm <leif.lindholm@linaro.org>:
> On Wed, Oct 11, 2017 at 05:40:47PM +0200, Marcin Wojtas wrote:
>> Instead of using hardcoded value in PcdSystemMemorySize PCD,
>> obtain DRAM size directly from SoC registers, which are filled
>> by firmware during early initialization stage.
>>
>> Contributed-under: TianoCore Contribution Agreement 1.1
>> Signed-off-by: Marcin Wojtas <mw@semihalf.com>
>> ---
>> Platform/Marvell/Armada/Library/Armada70x0Lib/Armada70x0LibMem.c | 111 +++++++++++++++++++-
>> Platform/Marvell/Armada/Library/Armada70x0Lib/Armada70x0LibMem.h | 49 +++++++++
>> 2 files changed, 159 insertions(+), 1 deletion(-)
>>
>> diff --git a/Platform/Marvell/Armada/Library/Armada70x0Lib/Armada70x0LibMem.c b/Platform/Marvell/Armada/Library/Armada70x0Lib/Armada70x0LibMem.c
>> index 2cb2e15..22cbe47 100644
>> --- a/Platform/Marvell/Armada/Library/Armada70x0Lib/Armada70x0LibMem.c
>> +++ b/Platform/Marvell/Armada/Library/Armada70x0Lib/Armada70x0LibMem.c
>> @@ -36,8 +36,11 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
>> #include <Library/ArmPlatformLib.h>
>> #include <Library/DebugLib.h>
>> #include <Library/HobLib.h>
>> +#include <Library/IoLib.h>
>> #include <Library/MemoryAllocationLib.h>
>>
>> +#include "Armada70x0LibMem.h"
>> +
>> // The total number of descriptors, including the final "end-of-table" descriptor.
>> #define MAX_VIRTUAL_MEMORY_MAP_DESCRIPTORS 16
>>
>> @@ -47,6 +50,105 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
>>
>> STATIC ARM_MEMORY_REGION_DESCRIPTOR VirtualMemoryTable[MAX_VIRTUAL_MEMORY_MAP_DESCRIPTORS];
>>
>> +// Obtain DRAM size basing on register values filled by early firmware.
>> +STATIC
>> +UINT64
>> +DramSizeGet (
>
> GetDramSize?
>
>> + UINT64 *MemSize
>
> IN, OUT?
>
2x OK.
>> + )
>> +{
>> + UINT64 BaseAddr;
>> + UINT32 RegVal;
>> + UINT8 AreaLengthMap;
>> + UINT8 Cs;
>> +
>> + *MemSize = 0;
>> +
>> + for (Cs = 0; Cs < DRAM_MAX_CS_NUM; Cs++) {
>> +
>> + RegVal = MmioRead32 (DRAM_CH0_MMAP_LOW_REG (Cs));
>> +
>> + /* Exit loop on first disabled DRAM CS */
>> + if (!(RegVal & DRAM_CS_VALID_ENABLED_MASK)) {
>> + break;
>> + }
>> +
>> + /*
>> + * Sanity check for base address of next DRAM block.
>> + * Only continuous space will be used.
>> + */
>> + BaseAddr = ((UINT64)MmioRead32 (DRAM_CH0_MMAP_HIGH_REG (Cs)) <<
>> + DRAM_START_ADDR_HTOL_OFFS) | (RegVal & DRAM_START_ADDRESS_L_MASK);
>
> Please use macros, temporary variables, more parentheses or whatever
> to help make the above operation readable.
>
Sure.
>> + if (BaseAddr != *MemSize) {
>> + DEBUG ((DEBUG_ERROR,
>> + "DramSizeGet: DRAM blocks are not contiguous, limit size to 0x%llx\n",
>> + *MemSize));
>> + return EFI_SUCCESS;
>> + }
>> +
>> + /* Decode area length for current CS from register value */
>> + AreaLengthMap = ((RegVal & DRAM_AREA_LENGTH_MASK) >> DRAM_AREA_LENGTH_OFFS);
>> + switch (AreaLengthMap) {
>
> Why Map?
>
I stuck to mv-ddr convention, but I agree, 'AreaLength' will be better.
>> + case 0x0:
>> + *MemSize += 0x18000000;
>> + break;
>> + case 0x1:
>> + *MemSize += 0x30000000;
>> + break;
>> + case 0x2:
>> + *MemSize += 0x60000000;
>> + break;
>> + case 0x3:
>> + *MemSize += 0xC0000000;
>> + break;
>
> The above ones look if not devoid of pattern, at least a bit
> unexpected - and 4-6 are comlpetely missing. Is there any
> documentation available for me to read to try to understand what is
> going on?
Do you have a docs for armada 7k or 8k? Please check 0xf0020200 table,
bits 20:16. You are right, I missed '4' (6GB), but 5 and 6 are
reserved (no value possible).
>
> The below all look predictably formatted and possible to calculate
> rather than list one by one.
>
> (1 << ((AreaLengthMap + 16))) if a quick calculation serves me right.
I think we can reduce the switch-case to this:
if (AreaLength =< 0x4) {
*MemSize = 0x18000000 << AreaLength;
} else if (AreaLength >= 0x7 && AreaLength < 0x1B)
*MemSize = 1 << (AreaLengthMap + 16);
} else {
DEBUG ((DEBUG_ERROR, "Invalid memory module size (0x%x) for
CS#%d\n", AreaLength, Cs));
}
Is above ok for you?
>
>> + case 0x7:
>> + *MemSize += 0x00800000;
>> + break;
>> + case 0x8:
>> + *MemSize += 0x01000000;
>> + break;
>> + case 0x9:
>> + *MemSize += 0x02000000;
>> + break;
>> + case 0xA:
>> + *MemSize += 0x04000000;
>> + break;
>> + case 0xB:
>> + *MemSize += 0x08000000;
>> + break;
>> + case 0xC:
>> + *MemSize += 0x10000000;
>> + break;
>> + case 0xD:
>> + *MemSize += 0x20000000;
>> + break;
>> + case 0xE:
>> + *MemSize += 0x40000000;
>> + break;
>> + case 0xF:
>> + *MemSize += 0x80000000;
>> + break;
>> + case 0x10:
>> + *MemSize += 0x100000000;
>> + break;
>> + case 0x11:
>> + *MemSize += 0x200000000;
>> + break;
>> + case 0x12:
>> + *MemSize += 0x400000000;
>> + break;
>> + case 0x13:
>> + *MemSize += 0x800000000;
>> + break;
>> + default:
>> + DEBUG ((DEBUG_ERROR, "Invalid area length (0x%x) for CS#%d\n", AreaLengthMap, Cs));
>
> Area length isn't really a helpful debug message.
> "memory module size"?
>
Ok.
Thanks,
Marcin
next prev parent reply other threads:[~2017-10-12 5:44 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-10-11 15:40 [platforms: PATCH 0/8] Armada 7k/8k - memory improvements Marcin Wojtas
2017-10-11 15:40 ` [platforms: PATCH 1/8] Marvell/Armada: Implement EFI_RNG_PROTOCOL driver for EIP76 TRNG Marcin Wojtas
2017-10-11 16:47 ` Leif Lindholm
2017-10-11 18:15 ` Ard Biesheuvel
2017-10-12 4:39 ` Marcin Wojtas
2017-10-12 10:24 ` Leif Lindholm
2017-10-11 15:40 ` [platforms: PATCH 2/8] Marvell/Armada: Increase preallocated memory region size Marcin Wojtas
2017-10-11 16:56 ` Leif Lindholm
2017-10-11 15:40 ` [platforms: PATCH 3/8] Marvell/Armada: Remove custom reset library residues Marcin Wojtas
2017-10-11 16:56 ` Leif Lindholm
2017-10-11 15:40 ` [platforms: PATCH 4/8] Marvell/Armada: Add support from DRAM remapping Marcin Wojtas
2017-10-11 17:08 ` Leif Lindholm
2017-10-11 18:18 ` Ard Biesheuvel
2017-10-12 4:58 ` Marcin Wojtas
2017-10-12 10:29 ` Leif Lindholm
2017-10-11 15:40 ` [platforms: PATCH 5/8] Marvell/Armada: Add MemoryInitPeiLib that reserves secure region Marcin Wojtas
2017-10-11 17:11 ` Leif Lindholm
2017-10-11 15:40 ` [platforms: PATCH 6/8] Marvell/Armada: Enable dynamic DRAM size detection Marcin Wojtas
2017-10-11 17:56 ` Leif Lindholm
2017-10-12 5:47 ` Marcin Wojtas [this message]
2017-10-12 10:50 ` Leif Lindholm
2017-10-12 10:58 ` Marcin Wojtas
2017-10-11 15:40 ` [platforms: PATCH 7/8] Marvell/Armada: Armada70x0Lib: Add support for 32-bit ARM Marcin Wojtas
2017-10-11 17:57 ` Leif Lindholm
2017-10-11 15:40 ` [platforms: PATCH 8/8] Marvell/Armada: Add 32-bit ARM support Marcin Wojtas
2017-10-11 17:58 ` Leif Lindholm
2017-10-11 18:20 ` Ard Biesheuvel
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-list from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAPv3WKepp4piZonQLk9mOgg6xa5nJ-Lib76M=ZRnfDv-RRJ0ww@mail.gmail.com' \
--to=devel@edk2.groups.io \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox