Message ID | 20130919194752.GA19975@quad.lixom.net |
---|---|
State | New |
Headers | show |
On Thu, 2013-09-19 at 12:47 -0700, Olof Johansson wrote: > > - A fix for build breakage in the MTD subsystem for some PXA devices. > David Woodhouse has this patch in his for-next branch but has not been > responding to our requests to send it up so here it is. > I should have amended the commit message to describe the build failure for > CONFIG_OF=n setups, but forgot and now it's down in the stack of commits. Thanks for pulling that in. Apologies for the screwup. We were were working on a better way to fix the litter of #ifdef CONFIG_OF which wasn't ready for the merge window so I just left this one out since it *appeared* to be entirely cosmetic.
On Thu, Sep 19, 2013 at 2:39 PM, Woodhouse, David <david.woodhouse@intel.com> wrote: > On Thu, 2013-09-19 at 12:47 -0700, Olof Johansson wrote: >> >> - A fix for build breakage in the MTD subsystem for some PXA devices. >> David Woodhouse has this patch in his for-next branch but has not been >> responding to our requests to send it up so here it is. >> I should have amended the commit message to describe the build failure for >> CONFIG_OF=n setups, but forgot and now it's down in the stack of commits. > > Thanks for pulling that in. Apologies for the screwup. No worries. You're a busy guy these days. > We were were working on a better way to fix the litter of #ifdef > CONFIG_OF which wasn't ready for the merge window so I just left this > one out since it *appeared* to be entirely cosmetic. Yeah, the real bug is a missing stub function in the #else case, but if the ifdef is all going away there's no point in adding it. The ifdeffery around the match tables is something that Mark Brown brought up yesterday as well. With ACPI_PTR() and of_match_ptr() we at least don't need the ifdefs in the struct device, but we do still need it for the match tables. -Olof
On Thu, Sep 19, 2013 at 2:39 PM, Woodhouse, David <david.woodhouse@intel.com> wrote: > On Thu, 2013-09-19 at 12:47 -0700, Olof Johansson wrote: >> >> - A fix for build breakage in the MTD subsystem for some PXA devices. >> David Woodhouse has this patch in his for-next branch but has not been >> responding to our requests to send it up so here it is. >> I should have amended the commit message to describe the build failure for >> CONFIG_OF=n setups, but forgot and now it's down in the stack of commits. > > Thanks for pulling that in. Apologies for the screwup. Speaking of such process issues: there's an outstanding patch for a (small) memory leak that was introduced in the nand_base.c ONFI code in 3.12-rc1. Can I expect you to attend to these sort of -rcX (where X > 1) issues? Or should I send Linus the patch myself? I don't really want it to wait around until 3.13-rc1 to go to -stable. It's in l2-mtd.git (will need cherry-picked out of there) and ready for the next linux-next: commit b2bdf43fcc1d440d8f4e1d5c8c59bf2ca76204df Author: Brian Norris <computersforpeace@gmail.com> Date: Mon Sep 16 17:59:20 2013 -0700 mtd: nand: fix memory leak in ONFI extended parameter page Brian
On Thu, 2013-09-19 at 15:57 -0700, Brian Norris wrote: > On Thu, Sep 19, 2013 at 2:39 PM, Woodhouse, David > <david.woodhouse@intel.com> wrote: > > On Thu, 2013-09-19 at 12:47 -0700, Olof Johansson wrote: > >> > >> - A fix for build breakage in the MTD subsystem for some PXA devices. > >> David Woodhouse has this patch in his for-next branch but has not been > >> responding to our requests to send it up so here it is. > >> I should have amended the commit message to describe the build failure for > >> CONFIG_OF=n setups, but forgot and now it's down in the stack of commits. > > > > Thanks for pulling that in. Apologies for the screwup. > > Speaking of such process issues: there's an outstanding patch for a > (small) memory leak that was introduced in the nand_base.c ONFI code > in 3.12-rc1. Can I expect you to attend to these sort of -rcX (where X > > 1) issues? Or should I send Linus the patch myself? I don't really > want it to wait around until 3.13-rc1 to go to -stable. I'm aware of it, and will get to it shortly. Was going to round it up along with the one that Olof just sent, either some time this week at Plumbers or when I get home. If you want to stick it in the linux-mtd.git tree before that and send Linus a pull request, I have no objection. Does that m25p80 'fix 4 byte addressing' fix for Micron not also want to go in?
On Thu, Sep 19, 2013 at 4:19 PM, Woodhouse, David <david.woodhouse@intel.com> wrote: > On Thu, 2013-09-19 at 15:57 -0700, Brian Norris wrote: >> On Thu, Sep 19, 2013 at 2:39 PM, Woodhouse, David >> <david.woodhouse@intel.com> wrote: >> > On Thu, 2013-09-19 at 12:47 -0700, Olof Johansson wrote: >> >> >> >> - A fix for build breakage in the MTD subsystem for some PXA devices. >> >> David Woodhouse has this patch in his for-next branch but has not been >> >> responding to our requests to send it up so here it is. >> >> I should have amended the commit message to describe the build failure for >> >> CONFIG_OF=n setups, but forgot and now it's down in the stack of commits. >> > >> > Thanks for pulling that in. Apologies for the screwup. >> >> Speaking of such process issues: there's an outstanding patch for a >> (small) memory leak that was introduced in the nand_base.c ONFI code >> in 3.12-rc1. Can I expect you to attend to these sort of -rcX (where X >> > 1) issues? Or should I send Linus the patch myself? I don't really >> want it to wait around until 3.13-rc1 to go to -stable. > > I'm aware of it, and will get to it shortly. Was going to round it up > along with the one that Olof just sent, either some time this week at > Plumbers or when I get home. > > If you want to stick it in the linux-mtd.git tree before that and send > Linus a pull request, I have no objection. If you're up to it, then you can take it. I'll take it myself if I feel it's getting too late. The problem is just when I can't differentiate your quiet awareness from oblivion (or obliviousness?). > Does that m25p80 'fix 4 byte addressing' fix for Micron not also want to > go in? I suppose the m25p80 one can go in as well. It's not so much a bugfix as augmenting my original patch to support more devices. But it is safe enough. Brian
On Thu, 2013-09-19 at 17:44 -0700, Brian Norris wrote: > > Does that m25p80 'fix 4 byte addressing' fix for Micron not also want to > > go in? > > I suppose the m25p80 one can go in as well. It's not so much a bugfix > as augmenting my original patch to support more devices. But it is > safe enough. You say 'support more devices', but AFAICT the code in -rc1 will already *attempt* to enable 4-byte addressing on those devices, and fail. (It looks like set_4byte() can return failure, but its return code isn't checked, and anyway it's only a SPI write so there's no feedback from the *chip* if it fails due to lack of WREN; it can only indicate failure if the SPI connection falls over.) So I'm inclined to think of it as a bug fix and include it.
On Fri, Sep 27, 2013 at 3:59 AM, Woodhouse, David <david.woodhouse@intel.com> wrote: > On Thu, 2013-09-19 at 17:44 -0700, Brian Norris wrote: >> > Does that m25p80 'fix 4 byte addressing' fix for Micron not also want to >> > go in? >> >> I suppose the m25p80 one can go in as well. It's not so much a bugfix >> as augmenting my original patch to support more devices. But it is >> safe enough. > > You say 'support more devices', but AFAICT the code in -rc1 will already > *attempt* to enable 4-byte addressing on those devices, and fail. As will the code in every release since 3.7-rc1 (commit 8da28681eb1430fb6715c7aef67001acfbbbcba5). > (It looks like set_4byte() can return failure, but its return code isn't > checked, and anyway it's only a SPI write so there's no feedback from > the *chip* if it fails due to lack of WREN; it can only indicate failure > if the SPI connection falls over.) > > So I'm inclined to think of it as a bug fix and include it. Sure, it's a bug fix. Just a long-standing bug. We were sending invalid commands to Micron devices, so no n25q256a devices could have possibly been supported properly before the 3.12-rc cycle [*]. I'm not sure who actually tested these flash way back in 3.7. I still would qualify them as "new devices" for 3.12, but that's just semantics now. Please, do include it. Brian [*] Under special circumstances they could have been supported. For instance, some n25q256a packages defaulted to 4-byte addressing mode.