How long did the fix take, and who has it?Fix dates, and who has it
A dated timeline drawn to scale from 18 August to 19 October 2026: the researcher's own dates, the fix in the kernel, the seven stable kernels and the CVE record of 15 October, the reports around them, and what seven distributors' pages said on the morning of 19 October. Every day count is derived. Hover, tap or focus any marker for the exact words and the source.
18 dates, in order, from 18 August 2026 to 19 October 2026
18 August 2026 (Tuesday): Vercel opens its challenge.
A public bug bounty for Vercel's Sandbox product. It is the earliest date in this timeline, so the axis starts here.
Vercel, announcement dated 18 August 2026: “Window: Tuesday, August 18 to Tuesday, September 1, 2026”
The axis runs 62 days, from 18 August to 19 October (derived).
Not stated: which reports arrived on which days, other than the one Vercel calls its most important.
20 August 2026 (Thursday): First report to Vercel.
The researcher's write up gives 20 August for its first report. Vercel gives day three of its challenge, and day three of a window that opened on 18 August is 20 August (derived).
Vercel, findings post dated 14 October 2026: “The program's most important report arrived on day three.” Vercel gives a day of its challenge, not a date. The date 20 August is the researcher's own.
2 days after the window opened (derived).
Not stated: Vercel's own date for the report, and what any report contained.
4 September 2026 (Friday): Private report to the kernel.
The write up says the root cause went to the kernel's security contact and the networking maintainers, with candidate patches. This is where the 41 day clock on this timeline starts, and it is the researcher's date.
The researcher's write up, table of research milestones on the researcher's own site: “kernel root cause privately to the kernel security contact and networking maintainers, with candidate patches” The researcher's own account.
15 days after the first report (derived). 41 days before the stable releases of 15 October (derived).
Not stated: a confirmation in a kernel archive. git.kernel.org and lore.kernel.org sit behind a bot check and were not read.
7 September 2026 (Monday): Issue reproduced.
The write up says a kernel engineer replied that the issue was reproduced and the candidate patches acknowledged.
The researcher's write up, table of research milestones on the researcher's own site: “replies that he reproduced the issue and acknowledges the candidate patches” The researcher's own account.
3 days after the private report (derived).
Not stated: the kernel side's own record of this exchange. The archives were not read.
22 September 2026 (Tuesday): Patches go public.
The write up puts public submissions between 19 and 22 September. The fix commit's Closes line links a netdev message whose identifier carries 22 September, which is the date drawn here.
The researcher's write up, table of research milestones on the researcher's own site: “Public netdev submissions and revisions” The researcher's own account.
kernel.org stable changelog, copy of the commit message, changelog for 7.2.11: The message identifier in the Closes line is dated 22 September. The netdev page itself was not read.
23 days before the CVE record and the stable releases (derived).
Not stated: the content of those posts. The netdev page sits behind a bot check and was not read.
25 September 2026 (Friday), 10:25 UTC: CVE ID reserved.
The kernel's CVE authority reserves CVE 2026 99999. Reserved means the number exists; the record was published on 15 October.
CVE.org record, dateReserved field, 25 September 2026, 10:25 UTC: Read from the CVE Services API on 19 October 2026.
20 days before publication (derived).
Not stated: who asked for the number, or why it was reserved on this day.
6 October 2026 (Tuesday), 03:06 UTC: The Register reports the claim.
The report cites a bug bounty award and a statement by a Vercel executive. It is a news report, not a kernel or vendor record.
The Register, article of 6 October 2026, 03:06 UTC: “We've confirmed a KVM 0day through our Vercel Sandbox bounty program.” A Vercel executive's words as the article quotes them. Secondary.
9 days before the CVE record and the stable releases (derived).
Not stated: what the fix team knew by then.
7 October 2026 (Wednesday), 22:42 UTC: Fix authored.
The mainline fix commit carries an author date of 7 October. The stable changelogs later copy its message.
GitHub mirror of the mainline kernel tree, commit author date, 7 October 2026, 22:42 UTC.
kernel.org stable changelog, copy of the commit message, changelog for 7.2.11: “as a defence in depth measure” “A more complete validation at input is forthcoming, but needs more work.” The commit's own words about its scope.
1 day after the first press report (derived). The commit changes 6 files, adding 44 lines and removing 12 (derived).
Not stated: whether the fuller validation the message calls forthcoming has landed.
8 October 2026 (Thursday), 18:03 UTC: Fix committed.
The commit's committer date is 8 October. The write up lists 7 to 8 October for the guard being posted and accepted.
GitHub mirror of the mainline kernel tree, commit committer date, 8 October 2026, 18:03 UTC.
The researcher's write up, table of research milestones on the researcher's own site: “The revised upstream guard is posted and accepted into net.git” The researcher's own account.
1 day after authoring (derived). 7 days before the stable releases (derived).
Not stated: the review discussion. The netdev and kernel archive pages were not read.
9 October 2026 (Friday), 04:40 UTC: Merged into mainline.
The GitHub mirror shows the networking pull for 7.3 rc7, with the fix, merged on 9 October at 04:40 UTC. The researcher's write up and SOC Prime say 14 October. This timeline uses 9 October: the mirror dates the 7.3 rc7 tag 12 October, before 14 October.
GitHub mirror of the mainline kernel tree, merge commit of the networking pull, 9 October 2026, 04:40 UTC: Read from the GitHub API on 19 October 2026.
The researcher's write up, table of research milestones on the researcher's own site: “merged into the mainline” “Oct 14th, 2026” The researcher's own date, which this timeline does not use.
SOC Prime, analysis page, secondary, read 19 October 2026: “on October 14”
3 days after the first press report (derived). 6 days before the stable releases (derived). 5 days earlier than the 14 October in the write up (derived).
Not stated: a kernel.org page for the merge. Those pages sit behind a bot check and were not read.
12 October 2026 (Monday), 03:04 UTC: 7.3 rc7 tagged.
The tag is dated 12 October on the GitHub mirror. The CVE record gives 7.3 rc7 as the mainline version with the fix.
GitHub mirror of the mainline kernel tree, tag of 7.3 rc7, 12 October 2026, 03:04 UTC: Read from the GitHub API on 19 October 2026.
3 days after the mainline merge (derived). 3 days before the stable releases (derived).
Not stated: a kernel.org copy of the tag. Those pages sit behind a bot check and were not read.
14 October 2026 (Wednesday): Vercel says fixes pending.
Vercel's findings post discloses the existence of the flaw, and says it does so because the flaw had been separately disclosed in public.
Vercel, findings post dated 14 October 2026: “Fixes are under private review and CVEs are pending” “it was separately disclosed in public”
1 day before the stable releases (derived).
Not stated: which fixes were under review, or any date for them.
15 October 2026 (Thursday), 09:56 to 12:19 UTC: Seven stable kernels and the CVE.
Seven stable and longterm kernels are released, with release commits from 09:56 to 12:16 UTC. The CVE record follows at 12:19 UTC, and the AWS, Red Hat and Ubuntu pages go up the same day.
Release commits, UTC: 6.12.113 09:56, 6.18.57 09:58, 7.2.11 10:00, 5.10.272 12:13, 5.15.223 12:14, 6.1.190 12:15, 6.6.159 12:16.
kernel.org stable changelog, release commits of 15 October 2026, UTC: “commit 6785011f8b16abf40b1e9d8b2e332232b3e68822 upstream” Each of the seven changelogs names the mainline commit as upstream.
CVE.org record, datePublished field, 15 October 2026, 12:19 UTC.
AWS, security bulletin of 15 October 2026, 17:30 UTC: “Publication Date: 10/15/2026 10:30 AM PDT”
The CVE record was published 2 minutes 54 seconds after the last release commit (derived). It was 41 days after the researcher's private report, 23 days after the netdev submission and 9 days after the first press report (all derived).
Not stated: when Red Hat, Ubuntu and SUSE expect to ship a fixed kernel.
15 October 2026 (Thursday), 19:25 UTC: A third party posts the write up.
A third party posts the write up's text to the oss security mailing list. The post says the flaw's existence was disclosed a few days earlier.
oss security list, post of 15 October 2026, 19:25 UTC: “The existence of this vulnerability was disclosed a few days ago on Twitter” A third party. Secondary.
7 hours 6 minutes after the CVE record (derived).
Not stated: the write up's own first publication date. Its page data carries none.
16 October 2026 (Friday), 15:49 and 17:15 UTC: No KEV entry.
CISA's Known Exploited Vulnerabilities catalogue, version 2026.10.16, has 1,748 entries and none for this CVE. CISA's own entry in the CVE record rates exploitation none. Oracle publishes UEK errata and Microsoft names a fix for Azure Linux 3.0.
CISA, KEV feed version 2026.10.16, released 16 October 2026, 17:15 UTC: 1,748 entries; this CVE is not among them.
CVE.org record, CISA entry inside the record, 16 October 2026, 15:49 UTC: Exploitation none. Automatable no. Technical impact total.
1 day after the CVE record (derived).
Not stated: whether any host has been attacked. No KEV entry does not show that none has.
17 October 2026 (Saturday), 11:14 UTC: CVE record updated.
The CVE.org record's last update is 17 October, two days after publication.
CVE.org record, dateUpdated field, 17 October 2026, 11:14 UTC.
2 days after publication (derived).
Not stated: what changed in the update. This timeline does not compare versions.
18 October 2026 (Sunday), 23:14 UTC: Linux 7.3 released.
kernel.org lists 7.3 as mainline, released 18 October. On the same date FIRST's EPSS model gave this CVE 0.211 per cent (derived from 0.00211). Red Hat's page was last changed the same day.
kernel.org, release list, 7.3 mainline, 18 October 2026: The 7.3 tag is dated 23:14 UTC on the GitHub mirror.
FIRST, EPSS page, read 19 October 2026: “estimates the probability that a published CVE will be exploited in the wild in the next 30 days” A model's estimate, not an observed attack.
3 days after the CVE record (derived).
Not stated: whether the EPSS value reflects any real attack. The model does not say.
19 October 2026 (Monday), 01:33 UTC: The Register's article and the morning reading.
The Register publishes at 01:33 UTC and says a fix is percolating into Debian, RHEL and Ubuntu. At the reading, 07:55 to 08:30 BST, Debian lists a fixed kernel and Red Hat and Ubuntu list none.
The Register, article of 19 October 2026, 01:33 UTC: “warned it might represent a new way to attack Linux” “It's percolating into major distros like Debian, RHEL, and Ubuntu” Secondary. The claim of a new way to attack is the researcher's; the article repeats it.
4 days after the stable releases and the CVE record (derived).
Not stated: any later change to the pages. They were read once, between 07:55 and 08:30 BST.
On the main axis these dates sit too close to tell apart, so the magnified strip draws the 14 days to scale. Each date has its own marker, and the two dates on 15 October are drawn side by side. 6 to 19 October is days 49 to 62 on this axis, 14 calendar days counting both ends (derived).
Intervals (derived)
- 41 days (derived), private report to the stable releases. 4 September, the researcher's date for the private report, to 15 October, the date of the seven stable kernels and the CVE record.
- 23 days (derived), public patches to the stable releases. 22 September, when the netdev message identifier is dated, to 15 October. If the researcher's account holds, the class of bug was in public review for that long before the CVE record.
- 9 days (derived), first press report to the stable releases. 6 October to 15 October. Along the way the fix commit was authored on 7 October, 1 day after the report (derived), and merged into mainline on 9 October, 3 days after it (derived).
- 4 days (derived), stable releases to this reading. 15 October to the Monday morning of 19 October. On the pages read, three of seven distributors still listed no fixed kernel.
Distributor pages as read, read 19 October 2026, 07:55 to 08:30 BST
3 of 7 list no fixed kernel, 42.9 per cent (derived)
| Distributor | State on 19 October | Not stated |
|---|---|---|
| Red Hat | No fixed kernel: 11 of 11 rows Affected. The product table covers RHEL 6 to 10, RHEL for NVIDIA 26 and OpenShift 4. Every row reads Affected. | a date for a fixed kernel. |
| Ubuntu | No fixed kernel: 7 of 8 releases needed. Seven releases of Ubuntu's kernel package read needed. The eighth, 14.04, reads ignored, because it is past standard support. Ubuntu's own priority is Medium. | a date for a fixed kernel, and any Ubuntu security notice. |
| SUSE | No fixed kernel: 167 of 167 rows Affected. The rows run across products under general support, under long term service pack support and past end of life. | a date for a fixed kernel. |
| Debian | Fixed kernel: security updates and sid. The tracker lists 6.1.190 for bookworm security, 6.12.113 for trixie security, and 7.2.11 for forky and sid. The base release versions of bookworm and trixie read vulnerable. | the advisory dates. The Debian list pages answered with a challenge and were not read. |
| Amazon Linux | Fixed kernel: 7 advisories. Amazon Linux 2 kernels 4.14, 5.4, 5.10 and 5.15, and Amazon Linux 2023 kernels 6.1, 6.12 and 6.18. All seven were updated on 15 October. Two show a 1 October release date. | when the fix entered each package. |
| Oracle UEK | Fixed kernel: errata dated 16 Oct. Three UEK errata, all dated 16 October, cover Oracle Linux 7 to 10. The page lists no erratum for the Red Hat Compatible Kernel. | an erratum for the Red Hat Compatible Kernel. |
| Azure Linux | Fixed kernel: fix named, 3.0. Microsoft's CSAF record names a fixed kernel for Azure Linux 3.0, first published 16 October at 08:07 UTC. It covers Azure Linux only. | anything outside Azure Linux, including Azure's hypervisor. |
One flaw, two scores
- 9.3 from Kernel CNA, privileges required: none. The only score on the CVE record itself. NVD holds the record at status Received and has no score of its own. Ubuntu sets its own priority at Medium. The vector recomputes to 9.3 (derived). The gap to 8.8 is 0.5 (derived), and privileges required is the only metric that differs.
- 8.8 from Red Hat, SUSE, Oracle, Amazon Linux, privileges required: low. The four publish the CNA's vector with privileges required set to Low. Red Hat says its metrics are preliminary. The pages read do not explain why they chose Low. The vector recomputes to 8.8 (derived). Both 9.3 and 8.8 are at or above Cyber Essentials' line of 7 (derived).
Exploitation is not established
- CISA, KEV feed version 2026.10.16, released 16 October 2026, 17:15 UTC: 1,748 entries. This CVE is not among them.
- CISA, entry inside the CVE record, 16 October 2026, 15:49 UTC: Exploitation none. Automatable no. Technical impact total.
- Red Hat, CVE page, read 19 October 2026: “Red Hat tracks this flaw as a KEV” The sentence goes on to say active exploitation. The KEV catalogue has no entry, and Red Hat's own data field for known exploitation reads false, so the line is unexplained.
- FIRST, EPSS page, read 19 October 2026: “estimates the probability that a published CVE will be exploited in the wild in the next 30 days” 0.211 per cent on 18 October (derived from 0.00211). A model, not an observed attack.
Not stated in the record
- Whether any host has been attacked.
- Whether a ready to run exploit exists. The write up says a second part is forthcoming.
- When Red Hat, Ubuntu and SUSE expect to ship a fixed kernel.
- Whether the fuller validation the commit calls forthcoming has landed.
- The dates of the Debian advisories. The list pages sit behind a challenge.
- Any statement from Google Cloud, Azure's hypervisor, Oracle Cloud, QEMU, Cloud Hypervisor, Proxmox or Nutanix.
Derived means this site's arithmetic. Ringed dates are the researcher's own account. No individual is named.
Derived means this site's arithmetic.