| Message ID | 20240325033515.814-1-jammy_huang@aspeedtech.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/ast: Fix soft lockup | expand |
Hi,
Thanks for your patch.
I'm wondering how you can trigger this infinite loop ?
Also this looks like a simple fix, that can be easily backported, so I'm
adding stable in Cc.
If Thomas has no objections, I can push it to drm-misc-fixes.
Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com>
Hi Am 27.03.24 um 09:53 schrieb Jocelyn Falempe: > Hi, > > Thanks for your patch. > I'm wondering how you can trigger this infinite loop ? Yeah, a bit more context for this bug would be welcome. It's hard to judge the fix without. Best regards Thomas > > Also this looks like a simple fix, that can be easily backported, so > I'm adding stable in Cc. > > If Thomas has no objections, I can push it to drm-misc-fixes. > > Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com> >
Hi Thomas and Jocelyn, What we do in ast_dp_set_on_off() is a handshake between host driver and bmc-fw to confirm the operation, on/off, is completed. We use some scratch registers in bmc to handshake with host. This handshake only work if BMC's scu-lock is opened. If scu-lock is opened too late, then it could lead to this issue. Best regards Jammy Thomas Zimmermann <tzimmermann@suse.de> 於 2024年3月28日 週四 上午4:28寫道： > Hi > > Am 27.03.24 um 09:53 schrieb Jocelyn Falempe: > > Hi, > > > > Thanks for your patch. > > I'm wondering how you can trigger this infinite loop ? > > Yeah, a bit more context for this bug would be welcome. It's hard to > judge the fix without. > > Best regards > Thomas > > > > > Also this looks like a simple fix, that can be easily backported, so > > I'm adding stable in Cc. > > > > If Thomas has no objections, I can push it to drm-misc-fixes. > > > > Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com> > > > > > > -- > -- > Thomas Zimmermann > Graphics Driver Developer > SUSE Software Solutions Germany GmbH > Frankenstrasse 146, 90461 Nuernberg, Germany > GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman > HRB 36809 (AG Nuernberg) > >
Hi Thomas and Jocelyn, What we do in ast_dp_set_on_off() is a handshake between host driver and bmc-fw to confirm the operation, on/off, is completed. We use some scratch registers in bmc to handshake with host. This handshake only work if BMC's scu-lock is opened. If scu-lock is opened too late, then it could lead to this issue. Best regards Jammy Thomas Zimmermann <tzimmermann@suse.de> 於 2024年3月28日 週四 上午4:28寫道： > > Hi > > Am 27.03.24 um 09:53 schrieb Jocelyn Falempe: > > Hi, > > > > Thanks for your patch. > > I'm wondering how you can trigger this infinite loop ? > > Yeah, a bit more context for this bug would be welcome. It's hard to > judge the fix without. > > Best regards > Thomas > > > > > Also this looks like a simple fix, that can be easily backported, so > > I'm adding stable in Cc. > > > > If Thomas has no objections, I can push it to drm-misc-fixes. > > > > Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com> > > > > > > -- > -- > Thomas Zimmermann > Graphics Driver Developer > SUSE Software Solutions Germany GmbH > Frankenstrasse 146, 90461 Nuernberg, Germany > GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman > HRB 36809 (AG Nuernberg) >
Hi Am 01.04.24 um 04:20 schrieb 黃立銘: > Hi Thomas and Jocelyn, > > What we do in ast_dp_set_on_off() is a handshake between host driver > and bmc-fw to confirm > the operation, on/off, is completed. > > We use some scratch registers in bmc to handshake with host. This > handshake only work if > BMC's scu-lock is opened. If scu-lock is opened too late, then it > could lead to this issue. Thanks a lot. Can you please send a new version of this patch with this information in the commit message? Please also mention why 200 ms is a good upper limit. The code currently waits and then possibly breaks the loop. Should the if-branch be located before the mdelay() statement to avoid any unnecessary waiting? Please also send the patch from your Aspeed email address. Our scripts do not accept patches where the sender differs from the Signed-off-by tag. Best regards Thomas > > Best regards > Jammy > > Thomas Zimmermann <tzimmermann@suse.de> 於 2024年3月28日 週四 上午4:28寫道： >> Hi >> >> Am 27.03.24 um 09:53 schrieb Jocelyn Falempe: >>> Hi, >>> >>> Thanks for your patch. >>> I'm wondering how you can trigger this infinite loop ? >> Yeah, a bit more context for this bug would be welcome. It's hard to >> judge the fix without. >> >> Best regards >> Thomas >> >>> Also this looks like a simple fix, that can be easily backported, so >>> I'm adding stable in Cc. >>> >>> If Thomas has no objections, I can push it to drm-misc-fixes. >>> >>> Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com> >>> >> >> >> -- >> -- >> Thomas Zimmermann >> Graphics Driver Developer >> SUSE Software Solutions Germany GmbH >> Frankenstrasse 146, 90461 Nuernberg, Germany >> GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman >> HRB 36809 (AG Nuernberg) >>
diff --git a/drivers/gpu/drm/ast/ast_dp.c b/drivers/gpu/drm/ast/ast_dp.c index ebb6d8ebd44e..1e9259416980 100644 --- a/drivers/gpu/drm/ast/ast_dp.c +++ b/drivers/gpu/drm/ast/ast_dp.c @@ -180,6 +180,7 @@ void ast_dp_set_on_off(struct drm_device *dev, bool on) { struct ast_device *ast = to_ast_device(dev); u8 video_on_off = on; + u32 i = 0; // Video On/Off ast_set_index_reg_mask(ast, AST_IO_VGACRI, 0xE3, (u8) ~AST_DP_VIDEO_ENABLE, on); @@ -192,6 +193,8 @@ void ast_dp_set_on_off(struct drm_device *dev, bool on) ASTDP_MIRROR_VIDEO_ENABLE) != video_on_off) { // wait 1 ms mdelay(1); + if (++i > 200) + break; } } }
Avoid infinite-loop in ast_dp_set_on_off(). Signed-off-by: Jammy Huang <jammy_huang@aspeedtech.com> --- drivers/gpu/drm/ast/ast_dp.c | 3 +++ 1 file changed, 3 insertions(+) base-commit: b0546776ad3f332e215cebc0b063ba4351971cca