| Summary: | [G35 KMS] regression confuses monitor to think resolution is 1920x1198 causing blur | ||
|---|---|---|---|
| Product: | DRI | Reporter: | Martin Bergstrom <marbe> |
| Component: | DRM/Intel | Assignee: | Chris Wilson <chris> |
| Status: | CLOSED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | medium | CC: | ben, chris, jbarnes |
| Version: | unspecified | ||
| Hardware: | x86 (IA32) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
| Attachments: | |||
Created attachment 49371 [details]
dmesg output under 2.6.32
Created attachment 49372 [details]
dmesg output under 2.6.39
Created attachment 49373 [details]
Various diagnostics under 2.6.32 (xrandr, fbset, ...)
Created attachment 49374 [details]
Various diagnostics under 2.6.39 (xrandr, fbset, ...)
Created attachment 49375 [details]
intel_reg_dumper output under 2.6.32
Created attachment 49376 [details]
intel_reg_dumper output under 2.6.39
Created attachment 49377 [details]
VBIOS dump
xrandr --newmode 1920p 154 1920 1968 2000 2080 1200 1203 1209 1235 +Hsync +VSync xrandr --addmode HDMI1 1920p xrandr --output HDMI1 --mode 1920p Just as a wild guess. (This is becoming a regular nuisance, does anybody know what is up with sync on SDVO?) (In reply to comment #8) > xrandr --newmode 1920p 154 1920 1968 2000 2080 1200 1203 1209 1235 +Hsync > +VSync > xrandr --addmode HDMI1 1920p > xrandr --output HDMI1 --mode 1920p Thanks, this workaround works (for X, but that's all I'm using). Because of "+VSync"? Bug still present under Linux kernel 3.0.0-1-686-pae (entered Debian testing today), which is expected I guess. The workaround in comment #7 still works. commit ba68e086223a5f149f37bf8692c8cdbf1b0ba3ef Author: Paulo Zanoni <paulo.r.zanoni@intel.com> Date: Fri Jan 6 19:45:34 2012 -0200 drm/i915/sdvo: always set positive sync polarity This is a revert of 81a14b46846fea0741902e8d8dfcc6c6c78154c8. We already set the mode polarity using the SDVO commands with struct intel_sdvo_dtd. We have at least 3 bugs that get fixed with this patch. The documentation, despite not clear, can also be interpreted in a way that suggests this patch is needed. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=15766 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=42174 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=43333 Cc: stable@kernel.org Reviewed-by: Eric Anholt <eric@anholt.net> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Keith Packard <keithp@keithp.com> Problem solved using the 3.2 Linux kernel that just entered Debian testing. (linux-image-3.2.0-1-686-pae, version 3.2.4-1) Thanks! Closing old verified. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
System environment: -- chipset: G35 -- system architecture: i686 -- kernel version: 2.6.32-5-686 vs 2.6.39-2-686-pae -- Linux distribution: Debian GNU/Linux Wheezy ("testing") -- Mobo model: ASUS P5E-V HDMI using onboard graphics -- Display connector: HDMI-converted-to-DVI I'm running Debian testing ("wheezy") and few months ago the Linux kernel was updated from 2.6.32 to 2.6.38 (and then 2.6.39). With this a regression occured causing some blurryness in two horizontal bands on my brand new EIZO FLEXSCAN S2433W-FS monitor. Under 2.6.39 the monitor OSD info says the resolution is 1920x1198, not 1920x1200 as it does with 2.6.32. The computer says the resolution is 1920x1200 under all kernel versions. I'm using the Intel graphics (driver i915) on the motherboard (ASUS P5E-V HDMI), and on its HDMI output there's a thingamajig converting HDMI to DVI (supplied with the motherboard), and from there a DVI-cable to the monitor. (Pretty sure it's DVI-D.) The motherboard also has an unused VGA-port. The blurriness is noticeable even before X starts, as soon as the kernel's KMS kicks in. I have done various diagnostics and nothing obvious stands out that I can see. The only things of note are: 1. xrandr calls the output "DVI1" under 2.6.32 and "HDMI1" under 2.6.39. 2. xrandr says "HDMI1 clones VGA1" under 2.6.38, nothing like that under 2.6.32. 3. xrandr mentions "force_audio: 0" under 2.6.38, not under 2.6.32. Trying to alter this makes not difference. 4. "fbset --show" says "accel true" under 2.6.38 but not under 2.6.32. Turning accel off make no difference. On the intel-gfx mailing list Chris Watson said this after looking at the output of intel_reg_dumper: > That narrows it down a lot. The modesetting registers are identical, so > the signal/clocks being sent to encoder are identical, the only > discrepancy there can be is then during the SDVO HDMI channel setup.