bugzilla-daemon at freedesktop.org
2012-Apr-20 09:25 UTC
[Nouveau] [Bug 48954] New: nv25 PGRAPH error and X freeze
https://bugs.freedesktop.org/show_bug.cgi?id=48954 Bug #: 48954 Summary: nv25 PGRAPH error and X freeze Classification: Unclassified Product: xorg Version: git Platform: x86 (IA32) OS/Version: Linux (All) Status: NEW Severity: normal Priority: medium Component: Driver/nouveau AssignedTo: nouveau at lists.freedesktop.org ReportedBy: noviktor at seznam.cz QAContact: xorg-team at lists.x.org Created attachment 60373 --> https://bugs.freedesktop.org/attachment.cgi?id=60373 Kernel log from boot to error on nv25 I am running nouveau git kernel 3.4.0-rc2-g965a47f, xf86-video-nouveau ver. fb3a36b and xorg-server 1.12.0 on nv25 card and I am getting PGRAPH erros nouveau 0000:01:00.0: PGRAPH - ERROR nsource: LIMIT_COLOR nstatus: PROTECTION_FAULT nouveau 0000:01:00.0: PGRAPH - ch 1/4 class 0x009f mthd 0x0308 data 0x00240771 X then freezes but ssh or SysRq works. Additionally some icons on Gnome 3 fall-back mode have white rectangle in background instead of being transparent. Didn't happen before libdrm rewrite. Fastest way to reproduce seems to be right clicking to open pop-up menu or opening drop-down selections - error happens in a few attempts. On nv11 it seems not to crash, but some icons still have white background. -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug.
Viktor Novotný
2012-May-26 20:15 UTC
[Nouveau] [PATCH] nv04/exa: Reset destination surface offset in the same call of NV04EXACopy
Fixes FDO bug #48954. Signed-off-by: Viktor Novotn? <noviktor at seznam.cz> --- src/nv04_exa.c | 15 ++++++++------- 1 files changed, 8 insertions(+), 7 deletions(-) diff --git a/src/nv04_exa.c b/src/nv04_exa.c index 7ede9d9..ca92868 100644 --- a/src/nv04_exa.c +++ b/src/nv04_exa.c @@ -220,7 +220,7 @@ NV04EXACopy(PixmapPtr pdpix, int srcX, int srcY, int dstX, int dstY, int split_dstY = NOUVEAU_ALIGN(dstY + 1, 64); int split_height = split_dstY - dstY; - if (nouveau_pushbuf_space(push, 16, 1, 0)) + if (nouveau_pushbuf_space(push, 16, 2, 0)) return; if ((width * height) >= 200000 && pNv->pspix != pNv->pdpix && @@ -249,7 +249,13 @@ NV04EXACopy(PixmapPtr pdpix, int srcX, int srcY, int dstX, int dstY, height -= split_height; dstY = 0; pNv->pmpix = pdpix; - } else + } + + BEGIN_NV04(push, NV01_BLIT(POINT_IN), 3); + PUSH_DATA (push, (srcY << 16) | srcX); + PUSH_DATA (push, (dstY << 16) | dstX); + PUSH_DATA (push, (height << 16) | width); + if (pNv->pmpix) { struct nouveau_bo *dst_bo = nouveau_pixmap_bo(pdpix); @@ -258,11 +264,6 @@ NV04EXACopy(PixmapPtr pdpix, int srcX, int srcY, int dstX, int dstY, pNv->pmpix = NULL; } - BEGIN_NV04(push, NV01_BLIT(POINT_IN), 3); - PUSH_DATA (push, (srcY << 16) | srcX); - PUSH_DATA (push, (dstY << 16) | dstX); - PUSH_DATA (push, (height << 16) | width); - if ((width * height) >= 512) PUSH_KICK(push); } -- 1.7.8.6
Marcin Slusarz
2012-Jun-07 12:51 UTC
[Nouveau] [PATCH] nv04/exa: Reset destination surface offset in the same call of NV04EXACopy
On Sat, May 26, 2012 at 10:15:20PM +0200, Viktor Novotn? wrote:> Fixes FDO bug #48954. > > Signed-off-by: Viktor Novotn? <noviktor at seznam.cz> > ---Reviewed-by: Marcin Slusarz <marcin.slusarz at gmail.com> It seems this code was buggy since it was introduced in "nv04-nv40/exa: Make the copy hook blit downwards for non-overlapping regions." (OFFSET_DESTIN was not zeroed for the second part), only "WIP: port to new libdrm" made it more apparent.> src/nv04_exa.c | 15 ++++++++------- > 1 files changed, 8 insertions(+), 7 deletions(-) > > diff --git a/src/nv04_exa.c b/src/nv04_exa.c > index 7ede9d9..ca92868 100644 > --- a/src/nv04_exa.c > +++ b/src/nv04_exa.c > @@ -220,7 +220,7 @@ NV04EXACopy(PixmapPtr pdpix, int srcX, int srcY, int dstX, int dstY, > int split_dstY = NOUVEAU_ALIGN(dstY + 1, 64); > int split_height = split_dstY - dstY; > > - if (nouveau_pushbuf_space(push, 16, 1, 0)) > + if (nouveau_pushbuf_space(push, 16, 2, 0)) > return; > > if ((width * height) >= 200000 && pNv->pspix != pNv->pdpix && > @@ -249,7 +249,13 @@ NV04EXACopy(PixmapPtr pdpix, int srcX, int srcY, int dstX, int dstY, > height -= split_height; > dstY = 0; > pNv->pmpix = pdpix; > - } else > + } > + > + BEGIN_NV04(push, NV01_BLIT(POINT_IN), 3); > + PUSH_DATA (push, (srcY << 16) | srcX); > + PUSH_DATA (push, (dstY << 16) | dstX); > + PUSH_DATA (push, (height << 16) | width); > + > if (pNv->pmpix) { > struct nouveau_bo *dst_bo = nouveau_pixmap_bo(pdpix); > > @@ -258,11 +264,6 @@ NV04EXACopy(PixmapPtr pdpix, int srcX, int srcY, int dstX, int dstY, > pNv->pmpix = NULL; > } > > - BEGIN_NV04(push, NV01_BLIT(POINT_IN), 3); > - PUSH_DATA (push, (srcY << 16) | srcX); > - PUSH_DATA (push, (dstY << 16) | dstX); > - PUSH_DATA (push, (height << 16) | width); > - > if ((width * height) >= 512) > PUSH_KICK(push); > } > --
bugzilla-daemon at freedesktop.org
2012-Jun-08 07:26 UTC
[Nouveau] [Bug 48954] nv25 PGRAPH error and X freeze
https://bugs.freedesktop.org/show_bug.cgi?id=48954 Ben Skeggs <skeggsb at gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution| |FIXED --- Comment #1 from Ben Skeggs <skeggsb at gmail.com> 2012-06-08 00:26:44 PDT --- I've pushed a patch to nouveau git which should fix this issue. -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug.
Reasonably Related Threads
- [PATCH] nv10/exa: Spring-cleaning
- [PATCH 1/2] Check before trying a solid fill
- [PATCH] nv04-nv40/exa: Reorder the commands in PrepareCopy to match the blob.
- [PATCH ddx] Add support for VRAM-less devices to the ddx
- [Bug 17377] New: NV50 failure on MacBook Pro.