linuxppc-dev.lists.ozlabs.org archive mirror
 help / color / mirror / Atom feed
From: Christoph Hellwig <hch@lst.de>
To: Geoff Levand <geoffrey.levand@am.sony.com>
Cc: linuxppc-dev@ozlabs.org, Paul Mackerras <paulus@samba.org>
Subject: Re: [PATCH 7/16] powerpc: add support for ps3 platform
Date: Wed, 15 Nov 2006 19:04:40 +0100	[thread overview]
Message-ID: <20061115180440.GB18856@lst.de> (raw)
In-Reply-To: <4554DACB.8060809@am.sony.com>

> Index: cell--common--6/arch/powerpc/platforms/ps3pf/mm.c
> ===================================================================
> --- /dev/null
> +++ cell--common--6/arch/powerpc/platforms/ps3pf/mm.c
> @@ -0,0 +1,872 @@
> +/*
> + * mm.c - PS3 Platform address space management.

Please never put the filename in the start of file comments, it
gets out of date far to easily.

> + *
> + *  Copyright (C) 2006 Sony Computer Entertainment Inc.
> + *  Copyright 2006 Sony Corp.

One with a (C) and one without looks very odd.  Who from your corporate
mother Sony Corp touched this code anyway?

> +#undef DEBUG

No need to put this in, it's undefined by default.  And having it
prevents people from enabling debugging through the compiler flags.

> +#if defined(DEBUG)
> +#undef pr_debug
> +#define pr_debug(fmt...) udbg_printf(fmt)
> +#endif

Please don't do this kind of thing.  Either you want pr_debug then
you use it as-is, or you want something different and you should call
it different.

> +
> +enum {
> +#if defined(CONFIG_PS3PF_USE_LPAR_ADDR)
> +	USE_LPAR_ADDR = 1,
> +#else
> +	USE_LPAR_ADDR = 0,
> +#endif
> +#if defined(CONFIG_PS3PF_DYNAMIC_DMA)
> +	USE_DYNAMIC_DMA = 1,
> +#else
> +	USE_DYNAMIC_DMA = 0,
> +#endif
> +};

This is really ugly.  At least dynamic_dma (or as we'd say it iommu)
shoud be a runtime feature.  So this should just be an

  "int ps3hv_use_iommu"
  
variable initialize to the proper default.

> +
> +enum page_size {
> +	page_size_4k = 12U,
> +	page_size_64k = 16U,
> +	page_size_16m = 24U,
> +};

Please use ALL_CAPS for such constants.


> +
> +enum allocate_memory {
> +	/* bit 63: transferability */
> +	LV1_AM_TF_NO = 0x00,
> +	LV1_AM_TF_YES = 0x01,
> +	/* bit 62: destruction scheme */
> +	LV1_AM_DS_NO_CONNECTIONS = 0x00,
> +	LV1_AM_DS_ANYTIME = 0x02,
> +	/* bit 61: fail or alternative */
> +	LV1_AM_FA_FAIL = 0x00,
> +	LV1_AM_FA_ALTERNATIVE = 0x04,
> +	/* bit 60: lpar address */
> +	LV1_AM_ADDR_ANY = 0x00,
> +	LV1_AM_ADDR_0 = 0x08,
> +};

If these are for different its they should be in different enums
and use different prefixes.

> +/**
> + * struct mem_region - memory region structure
> + * @base: base address
> + * @size: size in bytes
> + * @offset: difference between base and rm.size
> + */
> +
> +struct mem_region {
> +	unsigned long base;
> +	unsigned long size;
> +	unsigned long offset;
> +};

This seems like a duplication of the iseries lmb, please try to reuse
that functionality.

> + * @len: Length in bytes of the area to map.
> + * c_out: A pointer to receive an allocated struct dma_chunk for this area.
> + *
> + * This is the lowest level dma mapping routine, and is the one that will
> + * make the HV call to add the pages into the io controller address space.
> + */
> +
> +static int dma_map_pages(struct ps3pf_dma_region *r, unsigned long virt_addr,
> +	unsigned long len, struct dma_chunk **c_out)
> +{
> +	int result;
> +	unsigned long phys_addr;
> +	struct dma_chunk *c;
> +
> +	phys_addr = is_kernel_addr(virt_addr) ? __pa(virt_addr) : virt_addr;

When do you ever get non-kernel addresses into the dma calls? I in
either case this should move into the caller and the routine
should get a simple unsigned long phys_addr argument.

But when looking at this whole dma_map code is looks rather bogus to
me.  Could it be that you try to pre-map dma regions rather than doing
it through the dma API?

> +#if !defined(_510B7842_EE09_4B12_BE5C_D217383D50C7)
> +#define _510B7842_EE09_4B12_BE5C_D217383D50C7

WTF?

> +#if defined(CONFIG_BLK_DEV_INITRD)
> + 	if (initrd_start)
> + 		ROOT_DEV = Root_RAM0;
> + 	else
> +#endif
> +#if defined(CONFIG_ROOT_NFS)
> +		ROOT_DEV = Root_NFS;
> +#else
> + 		ROOT_DEV = Root_HDA1;
> +#endif

Please don't put this auto-select root device idiocy into any new
ports (Yeah, I'm pretty sure you just copy & pasted it from somewhere..)

> +#if !defined(PPC_MSG_MIGRATE_TASK)
> +static const int PPC_MSG_MIGRATE_TASK = 2;
> +#endif

This looks rather odd, what is it trying to do?

> +unsigned long __deprecated ps3pf_legacy_virq_to_outlet(unsigned int virq);

Never put deprecated routines into a new code submission.

  parent reply	other threads:[~2006-11-15 18:04 UTC|newest]

Thread overview: 35+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-11-10 20:02 [PATCH 7/16] powerpc: add support for ps3 platform Geoff Levand
2006-11-10 21:23 ` Olof Johansson
2006-11-14  0:11   ` Geoff Levand
2006-11-15 17:54     ` Christoph Hellwig
2006-11-21  2:50     ` Paul Mackerras
2006-11-21 10:11       ` Arnd Bergmann
2006-11-21 10:15         ` Geert Uytterhoeven
2006-11-21 11:59           ` Paul Mackerras
2006-11-21 13:22             ` Christoph Hellwig
2006-11-21 13:42               ` Arnd Bergmann
2006-11-21 13:45                 ` Christoph Hellwig
2006-11-21 19:54                   ` Geoff Levand
2006-11-21 19:54                 ` Geoff Levand
2006-11-21 19:54               ` Geoff Levand
2006-11-14 14:25   ` Geoff Levand
2006-11-11 11:21 ` Christoph Hellwig
2006-11-11 11:24   ` Christoph Hellwig
2006-11-11 11:50 ` Arnd Bergmann
2006-11-13  3:29   ` Geoff Levand
2006-11-14 14:00   ` Geoff Levand
2006-11-15 18:04 ` Christoph Hellwig [this message]
2006-11-16  1:41   ` Geoff Levand
2006-11-16  2:13     ` Michael Ellerman
2006-11-17  1:31       ` Geoff Levand
2006-11-17  6:41     ` Christoph Hellwig
2006-11-17 22:23       ` Benjamin Herrenschmidt
2006-11-18  2:02       ` Geoff Levand
2006-11-21 21:12       ` Geoff Levand
2006-11-17  0:39   ` Benjamin Herrenschmidt
2006-11-17  1:33     ` Geoff Levand
2006-11-17  1:42       ` Michael Ellerman
2006-11-17  2:04         ` Geoff Levand
2006-11-17  2:19           ` Michael Ellerman
2006-11-17  2:44             ` Geoff Levand
2006-11-16 16:35 ` Arnd Bergmann

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-all 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=20061115180440.GB18856@lst.de \
    --to=hch@lst.de \
    --cc=geoffrey.levand@am.sony.com \
    --cc=linuxppc-dev@ozlabs.org \
    --cc=paulus@samba.org \
    /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;
as well as URLs for NNTP newsgroup(s).