| Message ID | 20091124081827.6216.1896.stgit@angua (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Tue, 2009-11-24 at 01:18 -0700, Grant Likely wrote: > A cell is firmly established as a u32. No need to do an ugly typedef > to redefine it to cell_t. Eliminate the unnecessary typedef so that > it doesn't have to be added to the of_fdt header file > > Signed-off-by: Grant Likely <grant.likely@secretlab.ca> > --- I'm not sure about that one. Yes, we do use u32 a lot and cell_t rarely, so it would seem logical to switch.... On the other hand, we have that pesky endianness issue we have never fully solved. So we need accessors to sort that out, which means directly tapping things as u32 * is not a good idea if we're going to enforce the use of such accessors. I believe we should probably just enforce that properties are big endian for flat device-trees. In which case we could just use __be32 or on of thoes sparse-friendly types. I know x86 people won't like that much and to be honest I don't know what 1295 specifies for real OFs but there aren't enough real OFs around on LE machines for us to care much about it, is there ? The reason I prefer a fixed endianness is that allowing "LE" trees becomes really nasty when a number is expressed using multiple cells. That brings the question as to whether the two cells need to be flipped as well or only the bytes within each cell. And that's the easy bit (probably flip the whole thing). What about something like a PCI "reg" property which is made of 3 cells, two of them forming a 64-bit address and one containing additional data & attributes ? What is flipped and where ? So yes, cell_t might not be the right approach and by far to generic a name, but u32 isn't the answer neither. Cheers, Ben. > arch/microblaze/kernel/prom.c | 10 ++++------ > arch/powerpc/kernel/prom.c | 14 ++++++-------- > 2 files changed, 10 insertions(+), 14 deletions(-) > > diff --git a/arch/microblaze/kernel/prom.c b/arch/microblaze/kernel/prom.c > index e0f4c34..7760186 100644 > --- a/arch/microblaze/kernel/prom.c > +++ b/arch/microblaze/kernel/prom.c > @@ -42,8 +42,6 @@ > #include <asm/sections.h> > #include <asm/pci-bridge.h> > > -typedef u32 cell_t; > - > /* export that to outside world */ > struct device_node *of_chosen; > > @@ -159,7 +157,7 @@ static int __init early_init_dt_scan_memory(unsigned long node, > const char *uname, int depth, void *data) > { > char *type = of_get_flat_dt_prop(node, "device_type", NULL); > - cell_t *reg, *endp; > + u32 *reg, *endp; > unsigned long l; > > /* Look for the ibm,dynamic-reconfiguration-memory node */ > @@ -178,13 +176,13 @@ static int __init early_init_dt_scan_memory(unsigned long node, > } else if (strcmp(type, "memory") != 0) > return 0; > > - reg = (cell_t *)of_get_flat_dt_prop(node, "linux,usable-memory", &l); > + reg = (u32 *)of_get_flat_dt_prop(node, "linux,usable-memory", &l); > if (reg == NULL) > - reg = (cell_t *)of_get_flat_dt_prop(node, "reg", &l); > + reg = (u32 *)of_get_flat_dt_prop(node, "reg", &l); > if (reg == NULL) > return 0; > > - endp = reg + (l / sizeof(cell_t)); > + endp = reg + (l / sizeof(u32)); > > pr_debug("memory scan node %s, reg size %ld, data: %x %x %x %x,\n", > uname, l, reg[0], reg[1], reg[2], reg[3]); > diff --git a/arch/powerpc/kernel/prom.c b/arch/powerpc/kernel/prom.c > index 048e3a3..43cdba2 100644 > --- a/arch/powerpc/kernel/prom.c > +++ b/arch/powerpc/kernel/prom.c > @@ -67,8 +67,6 @@ int __initdata iommu_force_on; > unsigned long tce_alloc_start, tce_alloc_end; > #endif > > -typedef u32 cell_t; > - > extern rwlock_t devtree_lock; /* temporary while merging */ > > /* export that to outside world */ > @@ -441,22 +439,22 @@ static int __init early_init_dt_scan_chosen(unsigned long node, > */ > static int __init early_init_dt_scan_drconf_memory(unsigned long node) > { > - cell_t *dm, *ls, *usm; > + u32 *dm, *ls, *usm; > unsigned long l, n, flags; > u64 base, size, lmb_size; > unsigned int is_kexec_kdump = 0, rngs; > > ls = of_get_flat_dt_prop(node, "ibm,lmb-size", &l); > - if (ls == NULL || l < dt_root_size_cells * sizeof(cell_t)) > + if (ls == NULL || l < dt_root_size_cells * sizeof(u32)) > return 0; > lmb_size = dt_mem_next_cell(dt_root_size_cells, &ls); > > dm = of_get_flat_dt_prop(node, "ibm,dynamic-memory", &l); > - if (dm == NULL || l < sizeof(cell_t)) > + if (dm == NULL || l < sizeof(u32)) > return 0; > > n = *dm++; /* number of entries */ > - if (l < (n * (dt_root_addr_cells + 4) + 1) * sizeof(cell_t)) > + if (l < (n * (dt_root_addr_cells + 4) + 1) * sizeof(u32)) > return 0; > > /* check if this is a kexec/kdump kernel. */ > @@ -515,7 +513,7 @@ static int __init early_init_dt_scan_memory(unsigned long node, > const char *uname, int depth, void *data) > { > char *type = of_get_flat_dt_prop(node, "device_type", NULL); > - cell_t *reg, *endp; > + u32 *reg, *endp; > unsigned long l; > > /* Look for the ibm,dynamic-reconfiguration-memory node */ > @@ -540,7 +538,7 @@ static int __init early_init_dt_scan_memory(unsigned long node, > if (reg == NULL) > return 0; > > - endp = reg + (l / sizeof(cell_t)); > + endp = reg + (l / sizeof(u32)); > > DBG("memory scan node %s, reg size %ld, data: %x %x %x %x,\n", > uname, l, reg[0], reg[1], reg[2], reg[3]);
On Wed, Nov 25, 2009 at 8:59 PM, Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote: > On Tue, 2009-11-24 at 01:18 -0700, Grant Likely wrote: >> A cell is firmly established as a u32. No need to do an ugly typedef >> to redefine it to cell_t. Eliminate the unnecessary typedef so that >> it doesn't have to be added to the of_fdt header file >> >> Signed-off-by: Grant Likely <grant.likely@secretlab.ca> >> --- > > I'm not sure about that one. Yes, we do use u32 a lot and cell_t rarely, > so it would seem logical to switch.... On the other hand, we have that > pesky endianness issue we have never fully solved. So we need accessors > to sort that out, which means directly tapping things as u32 * is not a > good idea if we're going to enforce the use of such accessors. > > I believe we should probably just enforce that properties are big endian > for flat device-trees. In which case we could just use __be32 or on of > thoes sparse-friendly types. I know x86 people won't like that much and > to be honest I don't know what 1295 specifies for real OFs but there > aren't enough real OFs around on LE machines for us to care much about > it, is there ? Word from Mitch is the device tree is network byte order. period. > The reason I prefer a fixed endianness is that allowing "LE" trees > becomes really nasty when a number is expressed using multiple cells. > That brings the question as to whether the two cells need to be flipped > as well or only the bytes within each cell. And that's the easy bit > (probably flip the whole thing). What about something like a PCI "reg" > property which is made of 3 cells, two of them forming a 64-bit address > and one containing additional data & attributes ? What is flipped and > where ? exactly. > So yes, cell_t might not be the right approach and by far to generic a > name, but u32 isn't the answer neither. You're right, it's not, but makes merging less complex, and then I can refactor properly. g.
On Wed, 2009-11-25 at 21:05 -0700, Grant Likely wrote: > > You're right, it's not, but makes merging less complex, and then I can > refactor properly. Still, make them __be32 at least Cheers, Ben.
In message: <fa686aa40911252005o2db85dfk3d9acc61c12ca5e5@mail.gmail.com>
Grant Likely <grant.likely@secretlab.ca> writes:
: Word from Mitch is the device tree is network byte order. period.
OpenFirmware defines the order to be big endian always, even on little
endian processors.
Warner
On Wed, 2009-11-25 at 23:28 -0700, M. Warner Losh wrote: > In message: > <fa686aa40911252005o2db85dfk3d9acc61c12ca5e5@mail.gmail.com> > Grant Likely <grant.likely@secretlab.ca> writes: > : Word from Mitch is the device tree is network byte order. period. > > OpenFirmware defines the order to be big endian always, even on little > endian processors. Right, that's the only sane way to do it, I just didn't remember off hand what was said in the OF spec :-) Cheers, Ben.
> > Right, that's the only sane way to do it, I just didn't remember off > hand what was said in the OF spec :-) 3.2.2.1.2 Property values The property-encoding format is independent of hardware byte order and alignment characteristics. The encoded byte order is well-defined (in particular, it is big endian). ... ... -- 32-bit integer. A 32-bit integer is encoded into a property value byte array by storing the most significant byte at the next available address, followed (at address+1) by the high middle byte, the low middle byte, and (at address+3) the least significant byte.
>> You're right, it's not, but makes merging less complex, and then I >> can >> refactor properly. > > Still, make them __be32 at least There is no alignment guarantee at all either, better make it all u8 and use accessor functions everywhere. Segher
On Thu, 2009-11-26 at 22:36 +0100, Segher Boessenkool wrote: > >> You're right, it's not, but makes merging less complex, and then I > >> can > >> refactor properly. > > > > Still, make them __be32 at least > > There is no alignment guarantee at all either, better make it all u8 > and use accessor functions everywhere. Well... if you want to force using an accessor, then make it an opaque type. But __be32 is fine. It doesn't necessarily convey alignment and besides, there happens to -be- aligned in almost all cases so far :-) The flat tree format guarantees 32-bit alignment for the start of a property, so we are good here I think. Cheers, Ben.
From: Segher Boessenkool <segher@kernel.crashing.org> Date: Thu, 26 Nov 2009 22:36:41 +0100 >>> You're right, it's not, but makes merging less complex, and then I can >>> refactor properly. >> >> Still, make them __be32 at least > > There is no alignment guarantee at all either, better make it all u8 > and use accessor functions everywhere. I think that might be overkill.
On Wed, Nov 25, 2009 at 10:27 PM, Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote: > On Wed, 2009-11-25 at 21:05 -0700, Grant Likely wrote: >> >> You're right, it's not, but makes merging less complex, and then I can >> refactor properly. > > Still, make them __be32 at least Okay. Done. g.
diff --git a/arch/microblaze/kernel/prom.c b/arch/microblaze/kernel/prom.c index e0f4c34..7760186 100644 --- a/arch/microblaze/kernel/prom.c +++ b/arch/microblaze/kernel/prom.c @@ -42,8 +42,6 @@ #include <asm/sections.h> #include <asm/pci-bridge.h> -typedef u32 cell_t; - /* export that to outside world */ struct device_node *of_chosen; @@ -159,7 +157,7 @@ static int __init early_init_dt_scan_memory(unsigned long node, const char *uname, int depth, void *data) { char *type = of_get_flat_dt_prop(node, "device_type", NULL); - cell_t *reg, *endp; + u32 *reg, *endp; unsigned long l; /* Look for the ibm,dynamic-reconfiguration-memory node */ @@ -178,13 +176,13 @@ static int __init early_init_dt_scan_memory(unsigned long node, } else if (strcmp(type, "memory") != 0) return 0; - reg = (cell_t *)of_get_flat_dt_prop(node, "linux,usable-memory", &l); + reg = (u32 *)of_get_flat_dt_prop(node, "linux,usable-memory", &l); if (reg == NULL) - reg = (cell_t *)of_get_flat_dt_prop(node, "reg", &l); + reg = (u32 *)of_get_flat_dt_prop(node, "reg", &l); if (reg == NULL) return 0; - endp = reg + (l / sizeof(cell_t)); + endp = reg + (l / sizeof(u32)); pr_debug("memory scan node %s, reg size %ld, data: %x %x %x %x,\n", uname, l, reg[0], reg[1], reg[2], reg[3]); diff --git a/arch/powerpc/kernel/prom.c b/arch/powerpc/kernel/prom.c index 048e3a3..43cdba2 100644 --- a/arch/powerpc/kernel/prom.c +++ b/arch/powerpc/kernel/prom.c @@ -67,8 +67,6 @@ int __initdata iommu_force_on; unsigned long tce_alloc_start, tce_alloc_end; #endif -typedef u32 cell_t; - extern rwlock_t devtree_lock; /* temporary while merging */ /* export that to outside world */ @@ -441,22 +439,22 @@ static int __init early_init_dt_scan_chosen(unsigned long node, */ static int __init early_init_dt_scan_drconf_memory(unsigned long node) { - cell_t *dm, *ls, *usm; + u32 *dm, *ls, *usm; unsigned long l, n, flags; u64 base, size, lmb_size; unsigned int is_kexec_kdump = 0, rngs; ls = of_get_flat_dt_prop(node, "ibm,lmb-size", &l); - if (ls == NULL || l < dt_root_size_cells * sizeof(cell_t)) + if (ls == NULL || l < dt_root_size_cells * sizeof(u32)) return 0; lmb_size = dt_mem_next_cell(dt_root_size_cells, &ls); dm = of_get_flat_dt_prop(node, "ibm,dynamic-memory", &l); - if (dm == NULL || l < sizeof(cell_t)) + if (dm == NULL || l < sizeof(u32)) return 0; n = *dm++; /* number of entries */ - if (l < (n * (dt_root_addr_cells + 4) + 1) * sizeof(cell_t)) + if (l < (n * (dt_root_addr_cells + 4) + 1) * sizeof(u32)) return 0; /* check if this is a kexec/kdump kernel. */ @@ -515,7 +513,7 @@ static int __init early_init_dt_scan_memory(unsigned long node, const char *uname, int depth, void *data) { char *type = of_get_flat_dt_prop(node, "device_type", NULL); - cell_t *reg, *endp; + u32 *reg, *endp; unsigned long l; /* Look for the ibm,dynamic-reconfiguration-memory node */ @@ -540,7 +538,7 @@ static int __init early_init_dt_scan_memory(unsigned long node, if (reg == NULL) return 0; - endp = reg + (l / sizeof(cell_t)); + endp = reg + (l / sizeof(u32)); DBG("memory scan node %s, reg size %ld, data: %x %x %x %x,\n", uname, l, reg[0], reg[1], reg[2], reg[3]);
A cell is firmly established as a u32. No need to do an ugly typedef to redefine it to cell_t. Eliminate the unnecessary typedef so that it doesn't have to be added to the of_fdt header file Signed-off-by: Grant Likely <grant.likely@secretlab.ca> --- arch/microblaze/kernel/prom.c | 10 ++++------ arch/powerpc/kernel/prom.c | 14 ++++++-------- 2 files changed, 10 insertions(+), 14 deletions(-)