| Message ID | alpine.LFD.2.11.1501261318590.1322@knanqh.ubzr (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi! > > > $ du -b arch/arm/boot/dts/omap3-n900.dtb > > > 70212 arch/arm/boot/dts/omap3-n900.dtb > > > > > > $ echo $((0x10000)) > > > 65536 > > > > > > I would say, problem is because omap3-n900 binary DT is too large > > I agree. > > > OK if that's the case, then your patch makes sense to me. It also > > seems we can have the temporary stack be larger than the initial > > stack just for atags_to_fdt. > > The stack size isn't the issue, but rather its location. We need to > position it away from the DT data. The DT size is known and we could > use that, plus some room for the insertion of new data coming from the > ATAG conversion. > > Something like the following would be a more robust solution: > > diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S > index 68be901759..89718de9dd 100644 > --- a/arch/arm/boot/compressed/head.S > +++ b/arch/arm/boot/compressed/head.S > @@ -263,16 +263,37 @@ restart: adr r0, LC0 > * OK... Let's do some funky business here. > * If we do have a DTB appended to zImage, and we do have > * an ATAG list around, we want the later to be translated > - * and folded into the former here. To be on the safe side, > - * let's temporarily move the stack away into the malloc > - * area. No GOT fixup has occurred yet, but none of the > - * code we're about to call uses any global variable. > + * and folded into the former here. No GOT fixup has occurred > + * yet, but none of the code we're about to call uses any > + * global variable. > */ > - add sp, sp, #0x10000 > + > + /* Get the initial DTB size */ > + ldr r5, [r6, #4] > +#ifndef __ARMEB__ > + /* convert to little endian */ > + eor r1, r5, r5, ror #16 > + bic r1, r1, #0x00ff0000 > + mov r5, r5, ror #8 > + eor r5, r5, r1, lsr #8 > +#endif > + /* 50% DTB growth should be good enough */ > + add r5, r5, r5, lsr #1 > + /* preserve 64-bit alignment */ > + add r5, r5, #7 > + bic r5, r5, #7 > + /* clamp to 32KB min and 1MB max */ > + movs r1, r5, lsr #15 > + moveq r5, #(1 << 15) > + movs r1, r5, lsr #20 > + movne r5, #(1 << 20) Dunno. Would it be easier/simpler to just use 1MB, always? Do we support machines with <16MB RAM? Best regards,
On Mon, 26 Jan 2015, Pavel Machek wrote: > Hi! > > > > > $ du -b arch/arm/boot/dts/omap3-n900.dtb > > > > 70212 arch/arm/boot/dts/omap3-n900.dtb > > > > > > > > $ echo $((0x10000)) > > > > 65536 > > > > > > > > I would say, problem is because omap3-n900 binary DT is too large > > > > I agree. > > > > > OK if that's the case, then your patch makes sense to me. It also > > > seems we can have the temporary stack be larger than the initial > > > stack just for atags_to_fdt. > > > > The stack size isn't the issue, but rather its location. We need to > > position it away from the DT data. The DT size is known and we could > > use that, plus some room for the insertion of new data coming from the > > ATAG conversion. > > > > Something like the following would be a more robust solution: > > > > diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S > > index 68be901759..89718de9dd 100644 > > --- a/arch/arm/boot/compressed/head.S > > +++ b/arch/arm/boot/compressed/head.S > > @@ -263,16 +263,37 @@ restart: adr r0, LC0 > > * OK... Let's do some funky business here. > > * If we do have a DTB appended to zImage, and we do have > > * an ATAG list around, we want the later to be translated > > - * and folded into the former here. To be on the safe side, > > - * let's temporarily move the stack away into the malloc > > - * area. No GOT fixup has occurred yet, but none of the > > - * code we're about to call uses any global variable. > > + * and folded into the former here. No GOT fixup has occurred > > + * yet, but none of the code we're about to call uses any > > + * global variable. > > */ > > - add sp, sp, #0x10000 > > + > > + /* Get the initial DTB size */ > > + ldr r5, [r6, #4] > > +#ifndef __ARMEB__ > > + /* convert to little endian */ > > + eor r1, r5, r5, ror #16 > > + bic r1, r1, #0x00ff0000 > > + mov r5, r5, ror #8 > > + eor r5, r5, r1, lsr #8 > > +#endif > > + /* 50% DTB growth should be good enough */ > > + add r5, r5, r5, lsr #1 > > + /* preserve 64-bit alignment */ > > + add r5, r5, #7 > > + bic r5, r5, #7 > > + /* clamp to 32KB min and 1MB max */ > > + movs r1, r5, lsr #15 > > + moveq r5, #(1 << 15) > > + movs r1, r5, lsr #20 > > + movne r5, #(1 << 20) > > Dunno. Would it be easier/simpler to just use 1MB, always? Do we > support machines with <16MB RAM? If people are used to put other things relatively close to the kernel image like, say, some initrd image, then I'd prefer to be more conservative and avoid spreading out too much. Nicolas
> > > I would say, problem is because omap3-n900 binary DT is too large > > I agree. > > > OK if that's the case, then your patch makes sense to me. It also > > seems we can have the temporary stack be larger than the initial > > stack just for atags_to_fdt. > > The stack size isn't the issue, but rather its location. We need to > position it away from the DT data. The DT size is known and we could > use that, plus some room for the insertion of new data coming from the > ATAG conversion. > > Something like the following would be a more robust solution: Tested-by: Pavel Machek <pavel@ucw.cz> (Note, that in 3.19 dts for n900 got too big, so we are actually triggering old bugs. That means that this is a regression fix, and should go in ASAP). > > diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S > index 68be901759..89718de9dd 100644 > --- a/arch/arm/boot/compressed/head.S > +++ b/arch/arm/boot/compressed/head.S > @@ -263,16 +263,37 @@ restart: adr r0, LC0 > * OK... Let's do some funky business here. > * If we do have a DTB appended to zImage, and we do have > * an ATAG list around, we want the later to be translated > - * and folded into the former here. To be on the safe side, > - * let's temporarily move the stack away into the malloc > - * area. No GOT fixup has occurred yet, but none of the > - * code we're about to call uses any global variable. > + * and folded into the former here. No GOT fixup has occurred > + * yet, but none of the code we're about to call uses any > + * global variable. > */ > - add sp, sp, #0x10000 > + > + /* Get the initial DTB size */ > + ldr r5, [r6, #4] > +#ifndef __ARMEB__ > + /* convert to little endian */ > + eor r1, r5, r5, ror #16 > + bic r1, r1, #0x00ff0000 > + mov r5, r5, ror #8 > + eor r5, r5, r1, lsr #8 > +#endif > + /* 50% DTB growth should be good enough */ > + add r5, r5, r5, lsr #1 > + /* preserve 64-bit alignment */ > + add r5, r5, #7 > + bic r5, r5, #7 > + /* clamp to 32KB min and 1MB max */ > + movs r1, r5, lsr #15 > + moveq r5, #(1 << 15) > + movs r1, r5, lsr #20 > + movne r5, #(1 << 20) > + > + /* Temporarily relocate the stack past the DTB work space */ > + add sp, sp, r5 > stmfd sp!, {r0-r3, ip, lr} > mov r0, r8 > mov r1, r6 > - sub r2, sp, r6 > + mov r2, r5 > bl atags_to_fdt > > /* > @@ -285,11 +306,11 @@ restart: adr r0, LC0 > bic r0, r0, #1 > add r0, r0, #0x100 > mov r1, r6 > - sub r2, sp, r6 > + mov r2, r5 > bleq atags_to_fdt > > ldmfd sp!, {r0-r3, ip, lr} > - sub sp, sp, #0x10000 > + sub sp, sp, r5 > #endif > > mov r8, r6 @ use the appended device tree > @@ -306,7 +327,7 @@ restart: adr r0, LC0 > subs r1, r5, r1 > addhi r9, r9, r1 > > - /* Get the dtb's size */ > + /* Get the current DTB size */ > ldr r5, [r6, #4] > #ifndef __ARMEB__ > /* convert r5 (dtb size) to little endian */
On Tue, 27 Jan 2015, Pavel Machek wrote: > > > > > I would say, problem is because omap3-n900 binary DT is too large > > > > I agree. > > > > > OK if that's the case, then your patch makes sense to me. It also > > > seems we can have the temporary stack be larger than the initial > > > stack just for atags_to_fdt. > > > > The stack size isn't the issue, but rather its location. We need to > > position it away from the DT data. The DT size is known and we could > > use that, plus some room for the insertion of new data coming from the > > ATAG conversion. > > > > Something like the following would be a more robust solution: > > Tested-by: Pavel Machek <pavel@ucw.cz> > > (Note, that in 3.19 dts for n900 got too big, so we are actually > triggering old bugs. That means that this is a regression fix, and > should go in ASAP). It is queued here: http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8294/1 Nicolas
On Tuesday 27 January 2015 16:16:24 Nicolas Pitre wrote: > On Tue, 27 Jan 2015, Pavel Machek wrote: > > > > > I would say, problem is because omap3-n900 binary DT > > > > > is too large > > > > > > I agree. > > > > > > > OK if that's the case, then your patch makes sense to > > > > me. It also seems we can have the temporary stack be > > > > larger than the initial stack just for atags_to_fdt. > > > > > > The stack size isn't the issue, but rather its location. > > > We need to position it away from the DT data. The DT > > > size is known and we could use that, plus some room for > > > the insertion of new data coming from the ATAG > > > conversion. > > > > > Something like the following would be a more robust solution: > > Tested-by: Pavel Machek <pavel@ucw.cz> > > > > (Note, that in 3.19 dts for n900 got too big, so we are > > actually triggering old bugs. That means that this is a > > regression fix, and should go in ASAP). > > It is queued here: > http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id > =8294/1 > > > Nicolas I tested your patch and fdt_open_into() does not return error FDT_ERR_NOSPACE anymore. It is working. So you can add my Tested-by: Pali Rohár <pali.rohar@gmail.com>
* Nicolas Pitre <nicolas.pitre@linaro.org> [150127 07:19]: > On Tue, 27 Jan 2015, Pavel Machek wrote: > > > > > > > > I would say, problem is because omap3-n900 binary DT is too large > > > > > > I agree. > > > > > > > OK if that's the case, then your patch makes sense to me. It also > > > > seems we can have the temporary stack be larger than the initial > > > > stack just for atags_to_fdt. > > > > > > The stack size isn't the issue, but rather its location. We need to > > > position it away from the DT data. The DT size is known and we could > > > use that, plus some room for the insertion of new data coming from the > > > ATAG conversion. > > > > > > Something like the following would be a more robust solution: > > > > Tested-by: Pavel Machek <pavel@ucw.cz> > > > > (Note, that in 3.19 dts for n900 got too big, so we are actually > > triggering old bugs. That means that this is a regression fix, and > > should go in ASAP). > > It is queued here: > http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8294/1 Thanks for fixing that one! Regards, Tony
On Tue, Jan 27, 2015 at 10:16:24AM -0500, Nicolas Pitre wrote: > On Tue, 27 Jan 2015, Pavel Machek wrote: > > (Note, that in 3.19 dts for n900 got too big, so we are actually > > triggering old bugs. That means that this is a regression fix, and > > should go in ASAP). > > It is queued here: > http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8294/1 Irrespective of that patch, I commented on another thread (about adding the serial number to DT): | How about something a little more radical. | | Rather than trying to squeeze various ATAGs into DT, why don't we add a | standard ATAG to contain the DT and pass that through into the kernel. | This is IMHO how we _should_ have done the ATAG compatibility from the | start. | | That means we could get rid of most of the libfdt in the decompressor, | and instead resolve the differences in the kernel. That would allow us to get rid of all the FDT compatibility code in the decompressor, and when we encounter these special ATAGs, we can deal with them in the kernel instead of having to mess around with the DT. I'd also assume (although I haven't checked) that it would be much easier to add additional DT properties in the kernel environment, rather than in the rather restrictive decompressor environment. We'd need to be careful about how we deal with the atags vs DT. I'd suggest (as above) that we have a tag which points at the DT in physical memory, and its size, and pre-scan the atag list for that. When we find it, we kick off the DT based stuff as we would normally do, and follow it with a parse of the atags. This would cause the atags to override whatever is in the DT, which afaics is the behaviour that we actually want. We could augment that with code to update DT too so that we can export just the DT based information to kexec'd kernels.
On Tue, 27 Jan 2015, Russell King - ARM Linux wrote: > On Tue, Jan 27, 2015 at 10:16:24AM -0500, Nicolas Pitre wrote: > > On Tue, 27 Jan 2015, Pavel Machek wrote: > > > (Note, that in 3.19 dts for n900 got too big, so we are actually > > > triggering old bugs. That means that this is a regression fix, and > > > should go in ASAP). > > > > It is queued here: > > http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8294/1 > > Irrespective of that patch, I commented on another thread (about adding > the serial number to DT): > > | How about something a little more radical. > | > | Rather than trying to squeeze various ATAGs into DT, why don't we add a > | standard ATAG to contain the DT and pass that through into the kernel. > | This is IMHO how we _should_ have done the ATAG compatibility from the > | start. > | > | That means we could get rid of most of the libfdt in the decompressor, > | and instead resolve the differences in the kernel. > > That would allow us to get rid of all the FDT compatibility code in the > decompressor, and when we encounter these special ATAGs, we can deal with > them in the kernel instead of having to mess around with the DT. I'd > also assume (although I haven't checked) that it would be much easier to > add additional DT properties in the kernel environment, rather than in > the rather restrictive decompressor environment. Well... I was initially against any ATAG-to-DT compatibility at all. IMHO the "right" solution is to update the bootloader. However people complained that updating the bootloader wasn't possible. I therefore suggested they go with a shim layer I termed the "impedance matcher" acting like an intermediate boot stage converting their (often unofficial and out-of-tree) ATAGs into proper DT nodes. Some people did exactly that and it works wonderfully for them. It also works wonderfully for kernel maintenance as the platform specific tweaks are kept out of the kernel tree. Still, that wasn't good enough for some cases as the move to DT regressed some platforms that just used to work before. Concatenating a DTB to zImage was tolerable but installing an additional shim layer was too much. Hence the ATAG-to-DT compat code in the decompressor. This was meant to smooth things around the transition to DT, etc. After all, those devices with non-replaceable bootloaders where shim layers are not possible should get out of commission eventually? > We'd need to be careful about how we deal with the atags vs DT. I'd > suggest (as above) that we have a tag which points at the DT in > physical memory, and its size, and pre-scan the atag list for that. > When we find it, we kick off the DT based stuff as we would normally > do, and follow it with a parse of the atags. This would cause the > atags to override whatever is in the DT, which afaics is the behaviour > that we actually want. Maybe. However that would be yet another scheme for supporting legacy platforms with locked bootloaders. I really think we have enough of them already and I'd much prefer if we started _removing_ them not creating additional ones. The ATAG-to-DT code in the bootloader is still rather small and it works quite well, and it allows for configuring out ATAG support out of the kernel entirely. It should be sufficient for all the official (e.g. upstream) ATAGs already. If that's not good enough then people should consider the impedance matcher approach. > We could augment that with code to update DT too so that we can export > just the DT based information to kexec'd kernels. That should already be the case, no? Nicolas
On Tue, Jan 27, 2015 at 01:50:22PM -0500, Nicolas Pitre wrote: > Hence the ATAG-to-DT compat code in the decompressor. This > was meant to smooth things around the transition to DT, etc. After all, > those devices with non-replaceable bootloaders where shim layers are not > possible should get out of commission eventually? Probably after 5 years there are still people using and hacking mainline Linux with N900 etc., and newer OMAPs (since there no "useful" devices) are long forgotten... A.
On Tue 2015-01-27 13:50:22, Nicolas Pitre wrote: > On Tue, 27 Jan 2015, Russell King - ARM Linux wrote: > > > On Tue, Jan 27, 2015 at 10:16:24AM -0500, Nicolas Pitre wrote: > > > On Tue, 27 Jan 2015, Pavel Machek wrote: > > > > (Note, that in 3.19 dts for n900 got too big, so we are actually > > > > triggering old bugs. That means that this is a regression fix, and > > > > should go in ASAP). > > > > > > It is queued here: > > > http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8294/1 > > > > Irrespective of that patch, I commented on another thread (about adding > > the serial number to DT): > > > > | How about something a little more radical. > > | > > | Rather than trying to squeeze various ATAGs into DT, why don't we add a > > | standard ATAG to contain the DT and pass that through into the kernel. > > | This is IMHO how we _should_ have done the ATAG compatibility from the > > | start. > > | > > | That means we could get rid of most of the libfdt in the decompressor, > > | and instead resolve the differences in the kernel. > > > > That would allow us to get rid of all the FDT compatibility code in the > > decompressor, and when we encounter these special ATAGs, we can deal with > > them in the kernel instead of having to mess around with the DT. I'd > > also assume (although I haven't checked) that it would be much easier to > > add additional DT properties in the kernel environment, rather than in > > the rather restrictive decompressor environment. > > Well... > > I was initially against any ATAG-to-DT compatibility at all. IMHO the > "right" solution is to update the bootloader. > > However people complained that updating the bootloader wasn't possible. > I therefore suggested they go with a shim layer I termed the "impedance > matcher" acting like an intermediate boot stage converting their (often > unofficial and out-of-tree) ATAGs into proper DT nodes. Some people did > exactly that and it works wonderfully for them. It also works > wonderfully for kernel maintenance as the platform specific tweaks are > kept out of the kernel tree. > > Still, that wasn't good enough for some cases as the move to DT > regressed some platforms that just used to work before. Concatenating a > DTB to zImage was tolerable but installing an additional shim layer was > too much. Hence the ATAG-to-DT compat code in the decompressor. This > was meant to smooth things around the transition to DT, etc. After all, > those devices with non-replaceable bootloaders where shim layers are not > possible should get out of commission eventually? N900 is still phone that is best-supported by the mainline, AFAICT. Yes, it is rather sad, but it means that no, it is probably not going out of commission anytime soon. Pavel
On Tue 2015-01-27 10:16:24, Nicolas Pitre wrote: > On Tue, 27 Jan 2015, Pavel Machek wrote: > > > > > > > > I would say, problem is because omap3-n900 binary DT is too large > > > > > > I agree. > > > > > > > OK if that's the case, then your patch makes sense to me. It also > > > > seems we can have the temporary stack be larger than the initial > > > > stack just for atags_to_fdt. > > > > > > The stack size isn't the issue, but rather its location. We need to > > > position it away from the DT data. The DT size is known and we could > > > use that, plus some room for the insertion of new data coming from the > > > ATAG conversion. > > > > > > Something like the following would be a more robust solution: > > > > Tested-by: Pavel Machek <pavel@ucw.cz> > > > > (Note, that in 3.19 dts for n900 got too big, so we are actually > > triggering old bugs. That means that this is a regression fix, and > > should go in ASAP). > > It is queued here: > http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8294/1 Hmm, but it should be on kernel.org, not in private arm trees, as it fixes a regression. Russell, you are the ARM maintainer, can you push it to Linus? Pavel
On Mon, 2 Feb 2015, Pavel Machek wrote: > On Tue 2015-01-27 10:16:24, Nicolas Pitre wrote: > > On Tue, 27 Jan 2015, Pavel Machek wrote: > > > > > > > > > > > I would say, problem is because omap3-n900 binary DT is too large > > > > > > > > I agree. > > > > > > > > > OK if that's the case, then your patch makes sense to me. It also > > > > > seems we can have the temporary stack be larger than the initial > > > > > stack just for atags_to_fdt. > > > > > > > > The stack size isn't the issue, but rather its location. We need to > > > > position it away from the DT data. The DT size is known and we could > > > > use that, plus some room for the insertion of new data coming from the > > > > ATAG conversion. > > > > > > > > Something like the following would be a more robust solution: > > > > > > Tested-by: Pavel Machek <pavel@ucw.cz> > > > > > > (Note, that in 3.19 dts for n900 got too big, so we are actually > > > triggering old bugs. That means that this is a regression fix, and > > > should go in ASAP). > > > > It is queued here: > > http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8294/1 > > Hmm, but it should be on kernel.org, not in private arm trees, as it > fixes a regression. Russell, you are the ARM maintainer, can you push > it to Linus? The patch is included in RMK's "fixes" branch already, along with other fixes. I suppose it'll make its way to Linus before v3.19 final. Nicolas
diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S index 68be901759..89718de9dd 100644 --- a/arch/arm/boot/compressed/head.S +++ b/arch/arm/boot/compressed/head.S @@ -263,16 +263,37 @@ restart: adr r0, LC0 * OK... Let's do some funky business here. * If we do have a DTB appended to zImage, and we do have * an ATAG list around, we want the later to be translated - * and folded into the former here. To be on the safe side, - * let's temporarily move the stack away into the malloc - * area. No GOT fixup has occurred yet, but none of the - * code we're about to call uses any global variable. + * and folded into the former here. No GOT fixup has occurred + * yet, but none of the code we're about to call uses any + * global variable. */ - add sp, sp, #0x10000 + + /* Get the initial DTB size */ + ldr r5, [r6, #4] +#ifndef __ARMEB__ + /* convert to little endian */ + eor r1, r5, r5, ror #16 + bic r1, r1, #0x00ff0000 + mov r5, r5, ror #8 + eor r5, r5, r1, lsr #8 +#endif + /* 50% DTB growth should be good enough */ + add r5, r5, r5, lsr #1 + /* preserve 64-bit alignment */ + add r5, r5, #7 + bic r5, r5, #7 + /* clamp to 32KB min and 1MB max */ + movs r1, r5, lsr #15 + moveq r5, #(1 << 15) + movs r1, r5, lsr #20 + movne r5, #(1 << 20) + + /* Temporarily relocate the stack past the DTB work space */ + add sp, sp, r5 stmfd sp!, {r0-r3, ip, lr} mov r0, r8 mov r1, r6 - sub r2, sp, r6 + mov r2, r5 bl atags_to_fdt /* @@ -285,11 +306,11 @@ restart: adr r0, LC0 bic r0, r0, #1 add r0, r0, #0x100 mov r1, r6 - sub r2, sp, r6 + mov r2, r5 bleq atags_to_fdt ldmfd sp!, {r0-r3, ip, lr} - sub sp, sp, #0x10000 + sub sp, sp, r5 #endif mov r8, r6 @ use the appended device tree @@ -306,7 +327,7 @@ restart: adr r0, LC0 subs r1, r5, r1 addhi r9, r9, r1 - /* Get the dtb's size */ + /* Get the current DTB size */ ldr r5, [r6, #4] #ifndef __ARMEB__ /* convert r5 (dtb size) to little endian */